A common argument in favor of "don't ask, don't tell" — or of broader exclusion of homosexuals from military service — is that sexual tensions within a unit can seriously disrupt unit cohesion. I think that's a serious argument, which might indeed justify even what would otherwise be improper or even unconstitutional discrimination (though I stress the "might," since much depends on factual issues that I'm not confident about). And that's true not just about exclusion of gays and lesbians, but also about exclusion of women from combat units, something that indeed would be unconstitutional outside the military (though I agree there are other concerns supporting such exclusion as well, for instance the risk of sexual abuse of women who are taken prisoner by the enemy).
But what explains the exclusion of lesbians? Lesbians presumably serve in mixed-sex units already. Even if there are a few units that are all-women (which I highly doubt, except perhaps in traditionally female roles, such as nursing), they will certainly not be the sorts of combat units in which unit cohesion under fire is especially important. If anything, an openly lesbian woman in a mixed-sex unit is likely to create less sexual tension in that unit than a straight woman would (since the lesbian would likely have many fewer potential sexual partners within that unit than the straight woman).
One possible answer is that unit cohesion requires exclusion of homosexual men, and for reasons of sex equality the same rule needs to be applied to homosexual women. But that can't be right: After all, the military already discriminates based on sex in job assignments. If we tolerate sex-based exclusion of women (straight and gay) from combat roles, then why not equally tolerate sex-based inclusion of homosexual women even when there are overriding reasons why homosexual men need to be excluded?
Relatedly, what explains the exclusion of gay men from mixed-sex units, including lots of noncombat units? Again, an openly gay man in a mixed-sex unit is likely to create less sexual tension in that unit than a straight man or straight woman would. It's one thing to say "We have all-straight-male units, so the addition of anyone with whom sexual tension will be especially likely will undermine unit cohesion and military effectiveness." But once we allow women into the units, presumably on the view that sexual tension in those units doesn't much harm effectiveness (perhaps because sexual tension harms that effectiveness primarily in combat units), what's the unit cohesion rationale for nonetheless excluding lesbians or gays?
Now perhaps this reveals that the real rationale is something else, for instance the fear that many straight men and women will stop enlisting if they know there will be a few gays and lesbians around them (though surely most straight men and women must realize there will indeed be some gays and lesbians around, even if those gays and lesbians are in the closet), or concern about privacy in group showers and the like. I'm not persuaded by these rationales, but others might be. Still, if those are the real rationales for excluding lesbians (and excluding gays from mixed-sex units), and the unit cohesion rationale just doesn't apply to those particular exclusions in any material way, then it would be helpful to acknowledge this, shift the discussion to those rationales, and set the unit cohesion rationale aside in those contexts.
All this is especially important, by the way, given the practical costs of exclusion to military efficiency: The military spends a lot of time, effort, and money on training people. Discharging thousands of people who are otherwise effective soldiers, but who have somehow revealed their homosexuality, means a lot more spending on training replacements. And of course such discharges may well substantially hurt unit cohesion, as people lose their longtime colleagues with whom they've become friendly and with whom they've learned to work well.