Alvin L. Gilcrist entered a guilty plea to a charge of being a felon in possession of a firearm in violation of 18 U.S.C. § 922 (g)(1). Gilcrist was sentenced to 77 months, to be served consecutively to an undischarged state term.

Gilcrist challenges his sentence on two grounds. First, he takes issue with the district court’s computation of his criminal history category under the Guidelines. He argues that two of four prior convictions should not be counted under U.S.S.G. § 4A1.2(e)(l) because the periods of incarceration for each of them expired more than 15 years before the instant offense was committed.

Second, Gilcrist contends that application of the Guidelines in effect at the time he was sentenced violates the Ex Post Facto Clause of the Constitution. He argues that, absent a decision to depart upward, the Guidelines in force at the time he committed the instant offense foreclose a consecutive sentence, whereas the newer Guidelines do not foreclose a consecutive sentence, but instead grant discretion to the court to impose a consecutive sentence. We have jurisdiction under 28 U.S.C. § 1291 and 18 U.S.C. § 3742 .

We hold that two of Gilcrist’s prior sentences, for robbery and illegal possession of a weapon, should not have been included in calculating his criminal history category under U.S.S.G. § 4A1.2(e), and that the district court did not err in making Gilcrist’s sentence run consecutively under U.S.S.G. § 5G1.3(c). We also conclude that Gilcrist waived his ex post facto claim.

I. THE PRIOR CONVICTIONS

A. Background

The district court found it permissible to include all four of Gilcrist’s prior convictions in calculating his criminal history category under U.S.S.G. §§ 4Al.l(a) & 4A1.2(e)(l). Of these four convictions, Gilcrist contends that two, his conviction for robbery in 1972 and his conviction for possession of a weapon in 1974, were improperly considered. As relevant here, § 4A1.2(e)(l) provides that, to be counted under § 4Al.l(a), a prior sentence of imprisonment must have been

U.S.S.G. § 4A1.2(e)(l). Subsection (3) provides that any prior sentence “not within the time periods specified above is not counted.” U.S.S.G. § 4A1.2(e)(3). Gilcrist argues that his incarceration for robbery ended on March 17, 1975, the date of his parole for that offense, and that his incarceration for possession of a weapon ended on September 16, 1976, the date of his parole for that offense. Since the instant offense was committed on September 18, 1992, Gilcrist claims the district court erred in including those two convictions under § 4A1.2(e)(l).

The government responds by arguing that Gilcrist’s “parole” did not end his incarceration for those offenses. The government characterizes these paroles as nothing more than an administrative device to permit a prisoner to start serving time for a subsequent consecutive sentence. The government contends that this procedure is necessary because, in Washington, all sentences are consecutive, and in order to allow a prisoner to start serving a second sentence, he needs to be paroled from his previous sentence. Thus, the government contends, this procedure does not really end the prisoner’s incarceration for his previous crimes under § 4A1.2(e)(1).

B. Standard of Review

The relevant underlying facts are not disputed; the parties only disagree about the legal consequences of these facts under federal and state law. Therefore, the district court’s decision to count Gilerist’s robbery and weapon convictions under § 4A1.2(e) is reviewed de novo. United States v. Buenrostro-Torres, 24 F.3d 1173, 1174 (9th Cir.1994) (district court’s interpretation and application of Guidelines reviewed de novo)] United States v. Davis, 932 F.2d 752, 763 (9th Cir.1991) (determination whether prior conviction falls within the scope of Guidelines reviewed de novo).

C. Discussion

Gilcrist’s 1974 sentence for possession of a weapon was ordered to run consecutively to his 1972 sentence for robbery. On April 28, 1977, Gilcrist was paroled from the robbery sentence to the consecutive possession-of-a-weapon sentence. This parole was ordered to take effect, nunc pro tunc, on March 17, 1975. Similarly, in 1978, Gilcrist was paroled from the weapon sentence to a sentence imposed subsequently on a conviction for interference with a penal officer. Initially, the parole was set to take effect in 1978, but its effective date later became September 16, 1976, as a result of the application of certain legislation passed after the parole had been granted.

As the government concedes, under Washington law, Gilcrist had to wait until the expiration of incarceration under all prior terms to begin serving his interference sentence. Thus, when he began serving time on his interference sentence on September 16, 1976, it is inescapable that at that time his incarceration for robbery and weapons possession must have come to an end. In re Paschke, 61 WashApp. 591, 811 P.2d 694, 696 (1991) (“term of a subsequent felony sentence begins when the inmate’s actual imprisonment for the earlier felony ends”); Wash.Rev.Code § 9.92.080(1) (sentence for felony committed while under sentence of felony “shall not begin until the expiration of all prior terms”). Therefore, the district court erred in concluding that, under § 4A1.2(e), Gilcrist’s sentences for robbery and possession of a weapon “resulted in [him] being incarcerated” after September 16, 1976, and into the fifteen-year window, which started on September 18,1977.

It is undisputed that these paroles served the purpose of allowing Gilcrist to begin serving time on a consecutive sentence. However, whatever the motivation behind the state’s actions, its legal effect is clear: as of the start-date of the third consecutive sentence being served by Gilcrist (resulting from the interference conviction), he was no longer incarcerated on the two previous convictions. Otherwise, he would have started serving a consecutive sentence while still being incarcerated for a previous sentence, a legal impossibility under applicable Washington law. Wash.Rev.Code § 9.92.080(1); Paschke, 811 P.2d at 696 .

Gilcrist’s argument is further supported by the terms of his release on parole in May, 1991. Notwithstanding the government’s contention that he was still incarcerated for robbery at that point, Gilcrist was paroled not from his robbery sentence, but from his prison escape sentence. Indeed, the escape parole order recognized, as a simple reminder to Gilcrist, that he was “also on parole” for the robbery sentence. Had he not been previously paroled for robbery, he would not have been released from prison without an additional order of parole on his robbery conviction. As the Washington Department of Corrections acknowledged, by letter dated November 1, 1994, Gilcrist “concluded serving his time” for his robbery conviction on March 17,1975.

. Next, the government argues that In re Bible, 69 Wash.App. 394 , 845 P.2d 1336 (1992), requires a different conclusion. The government’s reliance on Bible is misplaced. That case held that the Parole Board has jurisdiction over a parolee until the expiration of his maximum term. This is an unremarkable proposition under Washington law and does not support the government’s argument that a parole to a consecutive sentence does not end the prisoner’s term of incarceration with respect to his previous sentence. Bible is consistent with the fact that, after such a parole to a consecutive cause, the prisoner is no longer incarcerated on account of his previous conviction, but is merely (like an ordinary parolee) subject to the Parole Board’s jurisdiction until the maximum term of his previous sentence expires. Id. 845 P.2d at 1338 .

The government also argues that Gil-crist’s legal interpretation of events would lead to anomalous results. Specifically, the government contends that under Gilerist’s theory, an unaccounted-for gap would exist between the time Gilcrist’s interference sentence ended on June 30, 1978, and the time his sentence for escape was imposed on July 27, 1978. According to the government, this “gap” is fatal to Gilcrist’s argument because had he in fact been paroled before June 30, 1978, he would have had to be released to the community during that period.

This argument lacks merit because, at the time the gap occurred, Gilcrist was incarcerated on another offense (assault), and it was impossible to predict at that time that this conviction would be overturned in 1990. Further, when the assault conviction was overturned in 1990, he was given credit against his subsequent escape sentence for that 28-day period, which the government says is unaccounted for, because during that time he actually was incarcerated awaiting sentencing following a conviction for escape. Wash. Rev.Code § 9.95.063 (time served during pendency of retrial is deducted from sentence imposed following retrial).

Finally, the government argues that because the paroles granted to Gilcrist did not involve a determination that he was rehabilitated and fit to be released to the community, he must have been incarcerated as a result of all of his outstanding convictions until the maximum terms of his sentences expired or until he actually was released to the community. We reject this argument. The reason he was paroled without a release determination was because he was expected, on the date of parole, to begin a term of incarceration for another sentence. Of course, if the subsequent sentence were eventually vacated or overturned, the Parole Board would have unconstrained authority to revoke parole, thus preventing the prisoner’s release. Wash.Rev.Code §§ 9.95.100 & 9.95.110 (Board “may return [parolee] to [prison] at its discretion”); Bible, 845 P.2d at 1339 .

Nonetheless, the government argues that, under § 9.95.100, the Parole Board is prohibited from paroling a prisoner until it finds that rehabilitation has been complete and that the prisoner is a fit subject for release. Thus, the government contends that the Board did not have authority to parole Gil-erist in the way it did. However, all that the statute prohibits is the release of a prisoner absent a determination that he is rehabilitated; it does not prohibit a parole to a consecutive sentence in the absence of a rehabilitation determination. In fact, this method of granting paroles to a consecutive sentence is supported fully by Washington law and has been routinely employed by the Parole Board for years. Wash.Admin.Code § 381-40-150 (authorizing “parole to consecutive sentence”); Paschke, 811 P.2d at 695-696 ; Jansen v. Morris, 87 Wash.2d 258 , 551 P.2d 743, 744 (1976) (en banc); Cook v. Rhay, 1 Wash. App. 931 , 465 P.2d 201, 202 (1970); St. Peter v. Rhay, 56 Wash.2d 297 , 352 P.2d 806 , 808 (1960) (en banc); In re Sanford, 10 Wash.2d 686 , 118 P.2d 179, 180 (1941). These paroles do not result from a meaningless administrative exercise, as the government seems to imply; on the contrary, they have a very significant legal effect in that they allow a prisoner to end incarceration for one sentence and start serving a subsequent term of imprisonment for a different and ^consecutive sentence.

On remand, the district court should resen-tence Gilcrist without including the robbery and weapons-possession convictions under § 4A1.2(e).

II. THE CONSECUTIVE SENTENCE

Gilcrist committed the instant offense while on parole from state charges. At the time of sentencing, his parole had been revoked and he was serving a 60-month parole revocation term. Consequently, applying U.S.S.G. § 5G1.3(c), comment, (n. 4), the district court decided Gilcrist’s sentence should run consecutively to his undischarged state parole revocation term.

Gilcrist challenges the court’s decision on two grounds: (i) application of the version of § 5G1.3(c) in force at the time of sentencing violated the Ex Post Facto Clause; and (ii) even if application of the latter version of § 5G1.3(e) were proper, a remand is required because the district court erroneously interpreted that provision to require the imposition of a consecutive sentence. We conclude that both contentions lack merit.

Gilcrist argues that the Guidelines in effect at the time he committed his offense, September, 1992, should apply. Under those Guidelines (the 1991 version), § 5G1.3(c) directed courts to use a particular methodology to determine whether a sentence should run consecutively. In the sense that it afforded courts less discretion, that version of § 5G1.3(c) was indeed more favorable to defendants than the one in force when Gilcrist was sentenced in 1995. United States v. Redman, 35 F.3d 437, 441 (9th Cir.1994), cert. denied, — U.S. —, 115 S.Ct. 922 , 130 L.Ed.2d 802 (1995).

Nonetheless, we cannot consider his ex post facto argument because Gilcrist, at the sentencing hearing, expressly agreed to the use of the Guidelines Manual in effect at the time of sentencing. Although he had asked the court, in his supplemental pro se sentencing brief, to use the 1991 Manual if it found the 1994 version to require a consecutive term, one week later, when asked at the sentencing hearing whether there were any objections to the use of the 1994 Manual, his defense counsel agreed to its use. In thus consenting, Gilcrist abandoned his ex post facto argument. See United States v. Warren, 980 F.2d 1300, 1304-06 (9th Cir.1992) (requiring use of only one version of the Guidelines in any given case).

Finally, Gilcrist argues. that application note four to § 5G1.3(c) does not require the imposition of a consecutive sentence, and that the district court erred because it did not realize it had discretion to impose a nonconsecutive sentence. We already have rejected this exact contention. United States v. Bernard, 48 F.3d 427, 431 (9th Cir.1995); accord United States v. Gondek, 65 F.3d 1, 2-3 (1st Cir.1995) (application note four is mandatory in spite of seemingly discretionary language).

III. CONCLUSION

Gilcrist’s sentence is vacated and remanded for resentencing. On remand, the district court shall omit Gilcrist’s two prior sentences for robbery and illegal possession of a weapon, occurring in 1972 and 1974 respectively, when calculating his criminal history category under U.S.S.G. § 4A1.2(e).

VACATED and REMANDED.

. The government also argues that, under federal concepts of "parole," Gilcrist was not "paroled" from the robbery or weapon convictions prior to September 18, 1977. However, this observation is irrelevant to whether, under § 4A1.2(e), the sentences for these offenses "resulted in [Gilcrist] being incarcerated” after that time. State law controls if, when, and for how long, Gilcrist is incarcerated for each of his state offenses. . The government argues that, even if the 28-day gap can be accounted for, a gap still would remain between the imposition of the suspended sentence for escape in July of 1978 and the time at which the suspension of his sentence was revoked in March of 1979. However, that period is accounted for as part of the time served for the escape sentence because the suspension was conditioned on Gilcrist serving 15 years on the assault charge, which condition was never met because his assault conviction was overturned and Gilcrist was acquitted of the assault charge on retrial in 1990. Therefore, as was recognized after the acquittal for assault, Gilcrist did legitimately serve time for his escape conviction between July of 1978 and March of 1979. . The district court is not foreclosed from considering whether an upward departure from the resulting criminal history category is warranted or permissible. See U.S.S.G. § 4A1.2, comment, (n. 8) (outlining conditions under which sentences too remote under § 4A1.2(e) may justify an upward departure under § 4A1.3); United States v. Beasley, 90 F.3d 400, 403 (9th Cir.) (same), cert. denied, —— U.S. —— , 117 S.Ct. 533 , 136 L.Ed.2d 418 (1996). . A district court’s interpretation and application of the Guidelines is reviewed de novo. See Part I.B., supra.