That isn't how genetics work. The first thing you've ignored is recessive genes. If a trait was always expressed in the phenotype we wouldn't have any fatal or extremely crippling genetic disorders that take hold before reproductive age, and the ones that take hold after reaching reproductive age would be much less common. But we do, because you can have and pass on a gene without expressing it or ever knowing you held it at all.
The second thing is that evolution is not rational. Evolution happens blindly and can go in utterly absurd directions. For example, sickle cell disease is most common among Africans. The reason for this is that while sickle cell can be crippling, it confers immunity to malaria, which is more crippling. If we were dealing with rational evolution the correct thing to do would just to evolve a cellular incompatibility to malaria that does not also involve decreasing the amount of oxygen it can carry. Something like that actually did happen in Europe in response to the Bubonic Plague, with 15% of modern Europeans carrying and 2% expressing a gene that confers natural immunity to Bubonic Plauge (and, as it so happens, HIV) and has no drawback.
Thirdly, an evolutionary adaptation only has to allow the species to reproduce and survive, not necessarily individuals. "Dead-end" traits can persist through carriers, and will if they confer some kind of benefit to the species. The "Gay Uncle" hypothesis suggests that homosexuality evolved as a kind of reinforcement upon our social structure to give more caretakers to children without risking an increase in the number of children. I don't really buy into that myself, but it is idea of how something like that can work.
Fourthly, human sexuality is extremely complex and by definition is poly-genetic, so there is definitely no "gay gene", but probably a set of genes that increase the likelihood of being attracted to your own sex or the other. Social impacts and pressures do the rest, so at the end of the day where you end up on the Kinsley Scale can't be attributed to genetics or society alone, but instead a myriad of factors.