[top posting modified]
i didnt imply that there is anything wrong necessarily with the idea of 'group selection' (see wilson/sober's interesting defense/history of it, in 'unto others'). just wanted to point out the trivial detail that most sociobiologists don't seem to want to concede any ground on their gene-centric approach.
there are so many different ways to go at the issue:
if the trait of homosexuality is genetically coded, as a recessive allele, then 1 of 4 (statistically) of the children of heterosexual parents (both with the recessive allele) has the chance of being homosexual. there is no reason to believe that there is a crushing fitness disadvantage to carrying the unexpressed (or perhaps semi-expressed) allele in such parents i.e., no reason to believe that the recessive allele will be wiped out of the population due to selection. perhaps there may even be reason to believe, if theories of the influence of recessive genes are correct, that it may increase fitness.
there is the possibility [which you seem to offer] that homosexuality has some fitness advantage, and with some breach of the sexual preference/attraction barrier, in order to carry out reproduction, the trait could have survived/thrived in the evolutionary record.
there is also the notion that we are all bisexual along a scale. some primates (bonobos?) have been shown to indulge in male homosexual behaviour to decrease incidents of communal violence -- surely a fitness advantage?
finally (not for the list of explanations, but for this post), there is the shockingly trivial idea that humans are extremely socio-cultural creatures and cultural evolution (social evolution?) must have kicked in very early in the evolutionary history of humankind. homosexuality could be one of those dawkinsian memes, an idea which has great survival fitness in the world of ideas. as the elaine character puts it in seinfeld: you can't compete against their team -- they have access to the equipment all day!
--ravi