MP-Ryan is correct. The case of overdominance or gender-antagonistic selection is probably not simply Mendelian. However, an allele that may promote fitness in one sex may contribute to homosexuality in the other (notice I said contribute, not determine!)
You argue that this means homosexuality is a 'byproduct', in which case my response is 'so what'? What's your point? It is a trait that boosts fitness. If homosexuality were a net disadvantageous side effect, it would have been selected against strongly, but it is not.
All this biological argument makes me slightly uncomfortable. The point here is simply to demonstrate that homosexuality is an important evolutionary phenomenon. Scientific explanations do not carry a moral component.
On the broader level, however, I don't think it matters. I would argue that there are absolutely no grounds for restricting the choices that consenting adults make so long as they do not harm each other or others...which these individuals do not.
Kara: good clarification of the point BengalTiger's trying to make. Sterility is indeed often selected for. Eusocial insects have sterile castes. Humans, however, do not have sterile castes. For sterility to be viable, we would need to be eusocial, and we are not. Homosexuals are therefore a better solution for kin selection behavior than sterile individuals, because the traits required to produce sterility would cripple heterosexuals more than they would produce useful kin helpers.