OPINION

WILLIAMS, Circuit Judge:

In this case, the Pepsi Cola Bottling Company of Fayetteville, Inc., (Pepsi), appeals from an order of the National Labor Relations Board (NLRB) ordering the payment of backpay to employees Christopher Hyatt and Robert Munn. Because Pepsi was prevented from introducing relevant evidence at the compliance proceeding which is necessary to evaluate the amount of Pepsi's backpay liability with respect to Hyatt, and because the NLRB has not explained adequately its use of a "representative employee" formula to calculate the amount of backpay due to Munn, we decline to enforce the NLRB's order, instead remanding for further proceedings.

I.

In the underlying unfair labor practices case, the NLRB's order required Pepsi, among other things, to reinstate and pay backpay to employees Hyatt and Munn. Pepsi-Cola Bottling Co. of Fayetteville, 315 NLRB 882 (1994) ( NLRB I). This portion of the NLRB's order was enforced by this Court in 1996. NLRB v. Pepsi-Cola Bottling Co. of Fayetteville, No. 95-1924, 1996 WL 511498, 1996 U.S.App. LEXIS 23936 (4th Cir. Sept. 10, 1996). On May 29, 1997, the NLRB's Regional Director issued a compliance specification, subsequently amended, setting the amount of backpay owed to Hyatt and Munn. In its answer, Pepsi challenged the compliance specification in several respects.

A hearing on the specification was held before an administrative law judge (ALJ) to resolve the issues presented. On February 9, 1998, the ALJ issued his supplemental decision, affirming the compliance specification for Hyatt, but setting aside the specification for Munn and recalculating the backpay owed to him. On exceptions from the ALJ's decision, the NLRB affirmed the ALJ's findings with respect to Hyatt, but reversed with respect to Munn, finding that the Regional Director's "representative employee" method of calculating the amount Munn would have earned had he remained at Pepsi should have been used and was erroneously rejected by the ALJ in favor of an average percentage pay increase approach. Pepsi-Cola Bottling Co. of Fayetteville, 330 NLRB No. 153 (2000) ( NLRB II).

A.

The backpay period for Hyatt began on December 30, 1992, when Hyatt was unlawfully discharged by Pepsi. Hyatt's backpay period, according to the NLRB, ended on May 12, 1997, when Pepsi reinstated him to his former position as a bulk route salesman. Between his discharge and reinstatement, Hyatt worked for several employers, including a Coca-Cola bottler and a firm known as D.K. Taylor. Pepsi asserts that after his reinstatement, Hyatt was terminated by Pepsi for failing a drug test, but the ALJ deemed this information irrelevant and did not permit inquiry into either Pepsi's drug testing policies or Hyatt's alleged failure of a Pepsi drug test.

In determining the amount of earnings Hyatt lost, the NLRB's Regional Director, in a finding upheld by the ALJ and the NLRB, applied a "representative employee" formula. Using information provided by Pepsi, the Regional Director initially identified a group of Pepsi's bulk route salesmen whose earnings approximated Hyatt's at the time of discharge. For each year of the backpay period, the average annual earnings of a representative group were calculated; these earnings were converted to a quarterly basis, and Hyatt was awarded the difference between his quarterly interim earnings during the period between his wrongful termination and reinstatement, and what he "would have" earned had he remained at Pepsi according to the representative employee calculation. (J.A. at 130.)

After his initial, wrongful termination from Pepsi, Hyatt took a job at Coca-Cola. In November of 1995, Hyatt resigned from his position at Coca-Cola after he failed a drug test. The NLRB asserts that Hyatt was, in fact, discharged, because, it posits, the only alternative to resignation was discharge. NLRB compliance officer Bradshaw stated that it "was [his] understanding" that Hyatt resigned involuntarily because he faced termination after failing a drug test. (S.A. at 46.) Bradshaw then stated that he was not competent to testify as to the circumstances surrounding Hyatt's separation from Coca-Cola and apologized for giving testimony regarding this issue. Hyatt himself was asked by Pepsi's attorney why he was "fired by Coke," and responded, "I failed a drug test." (J.A. at 58.) Then Hyatt clarified his testimony, stating that he did not supply the unemployment authorities with information regarding his separation from Coca-Cola "because actually I felt I wasn't terminated from any job that I left." (J.A. at 62.) After the ALJ summarized his testimony as being "that he wasn't terminated from any job, that he voluntarily quit," Hyatt responded, "Yes, sir." (J.A. at 62.) Asked again whether he voluntarily quit his Coca-Cola job, he again said, "Yes, sir." (J.A. at 61-62.)

After his departure from Coca-Cola, Hyatt was unemployed for approximately two weeks before accepting employment at a company known as D.K. Taylor at a lower pay rate than he earned at Coca-Cola but with more job responsibilities. The NLRB, in its backpay calculation, treated Hyatt as receiving earnings during his two-week period of unemployment at the rate he would have received had he remained at Coca-Cola. Beginning from Hyatt's start date at D.K. Taylor, however, the NLRB included as interim earnings only the actual earnings from Hyatt's lower-paying D.K. Taylor job. Thus, Pepsi's backpay liability reflects the reduction in earnings Hyatt suffered when he left Coca-Cola. Under the NLRB's calculation, the backpay owed Hyatt is $60,905.

B.

Robert Munn's backpay period began on November 22, 1991, when he was unlawfully discharged while he was still in his probationary period and earning $7.50 an hour as a mechanic. His backpay period ended on November 28, 1996, when Pepsi offered him reinstatement. As with Hyatt, the Regional Director, in an approach rejected by the ALJ but reinstated by the NLRB, applied a "representative employee" formula to determine the backpay owed Munn. The representative employee formula involved examining, on a quarterly basis, the earnings of mechanics considered comparable to Munn in terms of earnings at the time of their discharge; each quarter, the earnings of a pool of representative employees were averaged and the result was deemed to reflect what Munn would have earned had he remained at Pepsi. The NLRB determined that Munn was owed backpay of $18,871.64.

II.

The NLRB's interpretations of the Act are entitled to deference if they are reasonable, even if the NLRB's reading of the Act is not "the best way to read the statute." Holly Farms Corp. v. NLRB, 517 U.S. 392, 409, 116 S.Ct. 1396, 134 L.Ed.2d 593 (1996) (emphasis omitted). "If the [NLRB's] legal interpretations are `rational and consistent with the Act,' they will be upheld by reviewing courts." Sam's Club v. NLRB, 173 F.3d 233, 239 (4th Cir. 1999) (quoting Fall River Dyeing Finishing Corp. v. NLRB, 482 U.S. 27, 42, 107 S.Ct. 2225, 96 L.Ed.2d 22 (1987)) (internal citation omitted). When we review mixed questions [of fact and law], the [NLRB's] application of legitimate legal interpretations to the facts of a particular case should be upheld if they are supported by substantial evidence based upon the record as a whole. Sam's Club, 173 F.3d at 239. Substantial evidence review must consider evidence that detracts from as well as supports the NLRB's findings, and substantial evidence is "such relevant evidence as a reasonable mind might accept as adequate to support a conclusion." Id. (internal quotation marks omitted). A reviewing court engaged in substantial evidence review may not "displace the [NLRB's] choice between two fairly conflicting views" of the evidence, "even though the court would justifiably have made a different choice had the matter been before it de novo." Universal Camera Corp. v. NLRB, 340 U.S. 474, 488, 71 S.Ct. 456, 95 L.Ed. 456 (1951).

The Supreme Court has read 29 U.S.C.A. § 160(c) (West 1998) to "vest in the [NLRB] the primary responsibility and broad discretion to devise remedies that effectuate the policies of the Act, subject only to limited judicial review." Sure Tan, Inc. v. NLRB, 467 U.S. 883, 898-99, 104 S.Ct. 2803, 81 L.Ed.2d 732 (1984). "In fashioning its remedies under[the Act], . . . the [NLRB] draws on a fund of knowledge all its own, and its choice of remedy must therefore be given special respect by reviewing courts." NLRB v. Gissel Packing Co., 395 U.S. 575, 612 n. 32, 89 S.Ct. 1918, 23 L.Ed.2d 547 (1969). The NLRB's chosen remedy in a backpay case must be enforced "unless it is arbitrary, capricious, or manifestly contrary to the statute." Coronet Foods, Inc. v. NLRB, 158 F.3d 782, 788 (4th Cir. 1998) (internal quotation marks omitted). However, the findings of the NLRB are not to be "mechanically accepted" by the courts; instead, this Court is "obligated to scrutinize the whole record, taking into account whatever fairly detracts from the evidence relied upon by the [NLRB]." Id. at 789 (internal quotation marks omitted). Reviewing courts must "ensure that the [NLRB's] findings are not based on speculation or suspicion, as these register no weight on the substantial evidence scale." Id. (internal quotation marks omitted).

III.

Upon voluntarily resigning without good cause, an individual's entitlement to backpay is tolled because the individual has failed to mitigate damages. NLRB v. Hopcroft Art Stained Glass Works, Inc., 692 F.2d 63, 64 (8th Cir. 1982), citing Phelps Dodge Corp. v. NLRB, 313 U.S. 177, 61 S.Ct. 845, 85 L.Ed. 1271 (1941) (stating that "willful loss of earnings" disqualifies an individual from receiving backpay). Where an employee justifiably or reasonably quits interim employment, he does not forfeit his right to backpay. NLRB v. Ryder Sys., Inc., 983 F.2d 705 (6th Cir. 1993). The effect of an unjustified resignation is "that the amount the employee would have earned had he not quit is to be offset for the remainder of the backpay period." Hopcroft Art, 692 F.2d at 65. Where an employer "demonstrates that an employee did not exercise reasonable diligence in his or her efforts to secure interim employment, then it has established that the employee has not properly mitigated his or her damages." Tubari Ltd., Inc. v. NLRB, 959 F.2d 451, 454 (3d Cir. 1992). An employee who willfully loses employment by engaging in "deliberate or gross misconduct" is not entitled to backpay for a resulting earnings loss. Ryder System, Inc., 302 NLRB 608, 610 (1991), enforcement granted, 983 F.2d 705 (6th Cir. 1993).

A.

Pepsi argues, first, that Hyatt failed to mitigate his damages by immediately accepting a lower-paying position with D.K. Taylor rather than continuing to look for a higher-paying job, and thus, is entitled to no backpay whatsoever following his departure from Coca-Cola. Pepsi bases its argument on Tubari, which held that employees who immediately accepted employment as union picketers, at roughly 66% of their prior wages, after being wrongfully terminated, failed to demonstrate that they exercised reasonable diligence in locating suitable alternative employment and thus, were entitled to no backpay whatsoever. Tubari, 959 F.2d at 458-59. Pepsi's argument fails because it bears the burden of showing that Hyatt removed himself from the labor market, Lundy Packing Co. v. NLRB, 856 F.2d 627, 629 (4th Cir. 1988), and Pepsi developed no evidence that other, higher-paying jobs were available or that the job Hyatt accepted was not appropriate. Further, Tubari itself notes that "a discriminatee who accepts suitable alternative employment, even at a lower wage, has no continuing duty to search for a more lucrative job," demonstrating that the simple fact of lower pay does not suffice to preclude a finding that a discriminatee has found suitable alternative employment. Tubari, 959 F.2d at 458.

Additionally, Tubari did not involve a situation in which an employee initially found suitable alternative employment but then failed to exercise reasonable diligence in finding such employment after unjustifiably quitting his or her suitable alternative job. Complete preclusion from backpay eligibility may be appropriate in the case of an employee who never attempted to obtain a suitable job, because it is impossible to know how much the employee would have earned if he had diligently searched for an appropriate job. Here, however, we do know, in a sense, how much Hyatt would have earned had he diligently searched for suitable alternative employment, because we know how much he earned at Coca-Cola. We thus conclude that the rule in Hopcroft Art — that an employee who has unjustifiably quit is entitled to the difference between his prior earnings and the earnings he would have obtained had he not quit suitable alternative employment — should be invoked here. If Hyatt unjustifiably quit or was terminated for gross misconduct, he is entitled to the difference between his Pepsi and Coca-Cola earnings until the date of either his reinstatement at Pepsi or, as we discuss below, the date on which he would have been validly terminated by Pepsi, whichever is earlier.

B.

Next, Pepsi argues that no substantial evidence supports the NLRB's finding that Hyatt was functionally terminated because he was given the choice between termination and resignation. While the standard of review on this issue is quite deferential to the NLRB, this Court must take into account the evidence that contradicts, as well as evidence that supports, the NLRB's findings and must not mechanically affirm the NLRB's conclusions. Coronet Foods, Inc. v. NLRB, 158 F.3d 782, 788 (4th Cir. 1998). The NLRB argues that it was entitled to credit (1) the fact that when Hyatt was asked why he was "fired," he responded that he had failed a drug test, and (2) the testimony of a NLRB compliance officer that it was his "understanding" that Hyatt had been given a choice between quitting and being fired. We note that the NLRB's compliance officer testified in virtually the next breath that he was not competent to testify as to the circumstances surrounding Hyatt's resignation, but instead, he could testify only as to the method of calculating backpay, reducing significantly the probative value of his testimony on the issue of whether Hyatt was terminated or rather, voluntarily quit. As to Hyatt's statement, "because I failed a drug test," in response to a question asking him why he was "fired," it bears noting that a few pages later in Hyatt's testimony, he stated instead that he quit voluntarily and was not terminated by Coke. (S.A.61-62.) In light of Hyatt's conflicting testimony on this issue, it certainly cannot be said that the NLRB adduced strong evidence that Hyatt was, in effect, discharged by Coca-Cola. A reasonable trier of fact could conclude, however, that Hyatt was given an ultimatum to quit or face termination and that, in stating that he resigned "voluntarily," Hyatt simply meant that he could have forced Coca-Cola to terminate him by refusing to resign. Given that our review of the NLRB's factual findings is limited by the substantial evidence standard, we believe the NLRB's resolution of this factual issue was supported by "such relevant evidence as a reasonable mind might accept as adequate" to support the NLRB's conclusion. Sam's Club, 173 F.3d at 239.

Pepsi argues in the alternative that, even if Hyatt did not voluntarily quit, he was discharged for sufficiently egregious misconduct as to preclude the availability of enhanced backpay for his wage losses when he left Coca-Cola. The NLRB argues that Pepsi waived this issue by failing to raise it below as required by 29 U.S.C.A. § 160(e) (West 1998). The NLRB's claim of waiver is not well-founded. Pepsi asserted in its objection to the compliance specification that "[o]n information and belief . . . Hyatt was discharged from at least one interim employer because of his use of illegal drugs, thereby incurring a willful loss of interim earnings." (S.A. at 79.) The NLRB cites no authority in favor of its apparent assertion that the specific cases relied upon by Pepsi in support of this general proposition must have been cited to the NLRB to avoid waiver of this issue. While consideration of the NLRB's waiver argument is hampered by the fact that the Joint Appendix provided by the parties does not include Pepsi's list of exceptions to the NLRB from the ALJ's decision, an examination of the majority and dissenting NLRB opinions reveals that the NLRB itself explicitly split on the issue that the NLRB now claims was not presented to it. See NLRB II, 330 NLRB No. 153, 2000 WL 339963, at *4, *5 (2000).

In its Supplemental Decision, the NLRB determined that a discharge for any reason other than "moral turpitude" does not constitute a willful loss of earnings causing backpay to be calculated as if the employee had retained his prior job. Id. at *3. The NLRB then reasoned that because Pepsi presented no evidence surrounding the drug test incident other than Hyatt's failure of the test, the evidence did not establish moral turpitude. The NLRB's opinion leaves open the possibility that, for example, on-the-job intoxication that endangers coworkers could constitute "moral turpitude" but concludes that failing a drug test, without more, does not rise to the requisite level of misconduct. Id. Pepsi argues that this Court's decision in Brady v. Thurston Motor Lines, 753 F.2d 1269 (4th Cir. 1985), compels a finding that an employee who is discharged for cause "voluntarily remove[s] himself from the labor market . . . and forfeit[s] his right to back pay." Id. at 1277. Brady, however, was a Title VII backpay case that did not involve the deference to the NLRB that we must afford in this case, although it frequently cited NLRA case law as providing comparable principles. In Title VII cases, the courts set the backpay amount in the first instance, and no deference is due to any agency's method of calculating backpay. It cannot be said that the NLRB's "moral turpitude" standard is such an unreasonable approach, in light of the purposes of the Act, as to be reversible by this Court. Holly Farms Corp. v. NLRB, 517 U.S. 392, 409, 116 S.Ct. 1396, 134 L.Ed.2d 593 (1996) (stating that reasonable NLRB interpretations of the Act must be upheld even if they are not "the best way to read the statute" (emphasis omitted)). While the NLRB bears the burden of proof of showing the reasonableness of an employee's decision to voluntarily quit, see Minette Mills, Inc., 316 NLRB 1009, 1010 (1995), in general, the claim of a willful loss of earnings is an affirmative defense that the employer bears the burden of proving. See Lundy Packing Co., 856 F.2d at 629 (4th Cir. 1988); Florence Printing Co. v. NLRB, 376 F.2d 216, 223 (4th Cir. 1967). As a result, Pepsi bears the burden of proving that Hyatt's discharge was based upon gross misconduct. See Ryder System, Inc., 983 F.2d at 712. Pepsi had a full opportunity to question Hyatt regarding the circumstances surrounding his discharge from Coca-Cola and could have subpoenaed documents and witnesses to further develop the record regarding the circumstances of his discharge but produced no evidence other than the bare fact that Hyatt failed a drug test. The record does not reflect, for example, whether Hyatt used drugs in temporal proximity to his working hours, what kind of substances Hyatt may have consumed, or whether Hyatt's conduct ever evidenced the kind of workplace intoxication that could endanger himself or others. On these facts, the NLRB's conclusion that the record does not support a finding of moral turpitude is not unreasonable.

Pepsi argues further that the ALJ committed clear legal error in not allowing Pepsi to inquire into Hyatt's apparent failure of a drug test at Pepsi following his reinstatement, and to inquire generally into the nature of Pepsi's drug testing policies during the relevant period. It may be that Pepsi had a drug testing regime similar to that of Coca-Cola at the time Hyatt failed the Coca-Cola drug test; the ALJ's emphatic rejection of Pepsi's attempt to inquire into Hyatt's later alleged failure of a Pepsi drug test, following his reinstatement, clearly would be interpreted by a reasonable litigant as foreclosing such a line of inquiry. Yet information regarding Pepsi's drug testing program was of great relevance to this case. It is clear that an individual's backpay entitlement ends when that individual would have otherwise been terminated from employment in a legally permissible manner. Dockbuilders Local Union No. 1456, 316 NLRB 257, 259 (1995) (stating that the backpay period ends when employment would have been validly terminated due to an economic layoff). Thus, if at the time he failed the Coca-Cola drug test, Hyatt would have failed the Pepsi drug test, and if Pepsi had a policy of terminating individuals who fail a drug test, Hyatt would have been properly terminated by Pepsi on that date, and Pepsi's backpay liability should end at the time he would have been legitimately terminated from Pepsi. We thus conclude that it is necessary to remand for additional factual findings regarding the nature of Pepsi's drug testing policy during the relevant period.

IV.

The NLRB's choice of a method for calculating backpay is reviewed for substantial evidence. Coronet Foods, Inc. v. NLRB, 158 F.3d 782, 789-800 (4th Cir. 1998). Put differently, the NLRB's method of calculating backpay must be upheld unless "the method selected [is] arbitrary or unreasonable in the circumstances involved." Woodline Motor Freight, Inc. v. NLRB, 972 F.2d 222, 225 (8th Cir. 1992) (internal quotation marks omitted). The NLRB's General Counsel submitted a compliance specification that calculated Munn's backpay amount on the basis of a "representative employee" formula. The ALJ rejected the General Counsel's approach, finding that the representative employee formula led to a conclusion that at the end of his probationary period, Munn would have received a 16.7% pay increase, which the ALJ deemed excessive in light of the evidence; the ALJ concluded that Munn would not have received a pay increase of this magnitude at the end of his probationary period. Indeed, the ALJ noted that this increase would have resulted in Munn earning more than either his supervisor or other employees who had worked at Pepsi for a longer period of time in the same job. NLRB II, 330 NLRB No. 153, 2000 WL 339963, at *10. The NLRB reversed the ALJ without challenging any of these factual findings, stating simply that the representative employee formula "has long been used by the [NLRB]," and thus, the ALJ's decision to use a set annual average wage increase approach rather than a representative employee approach "was neither justified nor necessary." Id. at *2. The NLRB did not respond to the ALJ's legal conclusion, derived from the NLRB's precedents, that where there is a dispute as to the proper backpay formula, the ALJ must choose the "most accurate" formula, e.g. the formula that best captures the likely actual amount of wages lost due to illegal termination. See Coronet Foods, 158 F.3d at 800; Woodline Motor Freight, 305 NLRB 6 (1991). Nor did the NLRB find that the representative employee formula was more accurate than the annual-percentage approach, which the ALJ found to be the most accurate approach. Moreover, the NLRB has in the past acknowledged that it is appropriate to reject backpay formulas that are internally consistent but unreflective of the realities of the events involved. East Wind Enter., 268 NLRB 655, 656 (1984).

The background legal principle against which we must review the NLRB's approach is that "the back pay provisions of the NLRA . . . are compensatory and remedial in purpose, not punitive." Brady v. Thurston Motor Lines, 753 F.2d 1269, 1278 (4th Cir. 1985). Thus, while the NLRB's choice of a method for calculating backpay concerns a matter uniquely within the NLRB's competence and should be given a wide berth by reviewing courts, an explicitly punitive method of calculation is contrary to the purposes of the Act; backpay is not a form of punitive damages. If the NLRB had determined, based upon substantial evidence, that a representative employee formula would in this case be a more accurate method of calculating Munn's backpay, we would be bound to uphold that finding. As it is, however, the NLRB has upheld all ALJ findings not explicitly reversed — including the ALJ's straightforward finding that the pay increases contemplated by the representative employee formula exceeded those which Munn actually would have received had he remained at Pepsi. Thus, the NLRB may have decided that the representative employee approach overcompensates Munn and is less accurate than a percentage increase approach, but should be used nevertheless, because it has "long been used," even though more than a decade of NLRB precedent indicates that it should not be used when it is less accurate than another alternative. We must, in these circumstances, deny enforcement of this portion of the NLRB's order and remand for the NLRB to clarify the apparently irreconcilable tension between the NLRB's decision to implicitly affirm the ALJ's factual finding that the representative employee approach produced less accurate results and the NLRB's determination that this approach should be used to calculate backpay.

V.

In sum, we affirm the NLRB's findings that Hyatt was constructively discharged by Coca-Cola and that Pepsi did not carry its burden of showing that his constructive discharge was caused by an offense involving moral turpitude. We further affirm the NLRB's rejection of Pepsi's contention that Hyatt voluntarily removed himself from the labor market by taking a lower-paying job after his separation from Coca-Cola. We deny enforcement on the present record and remand, however, for further development of the record regarding Pepsi's drug testing policies, and the circumstances of Hyatt's failure of the Coca-Cola drug test insofar as they bear on the question of whether Hyatt would have failed a Pepsi drug test and been terminated by Pepsi for this reason at some point had he not been terminated earlier by Pepsi in violation of the Act. We further deny enforcement on the record before us and remand the calculation of Munn's backpay amount to the NLRB for reconsideration, so that the NLRB can resolve the apparent tension between its use of the representative employee formula and the lack of any finding that this formula is more accurate than the alternative used by the ALJ in this case.

REMANDED WITH INSTRUCTIONS.