We've known for quite a while that people, women in particular, tend to be attracted to men with different HLA profiles than theirs. This is particularly true when they are close to ovulation. When they are pregnant their preference changes to those with very similar profiles. The mechanism for this selection is smell.

This effect is particularly pronounced in groups like the Amish and Chassidim who practice severe inbreeding. Any shred of heterozygocity is a matter of group survival. It is also a reason why from a purely biological point of view the best thing that could happen to these groups would be successive invasions by the Mongol hordes, Vikings and Dayak pirates.

I hesitate to ascribe "reasons" to evolutionary phenomena, but a significant effect is that children sired by such men will have more diverse defenses against disease. It may also be a proxy for increased heterozygocity in general which confers many advantages. The second part, prefering the smell of close kin during pregnancy, may well be an artifact of our evolutionary past when we practiced female philopatrie - women sticking close to the family groups they were born into as many mammals do. Staying with close kin helped insure survival during pregnancy.

It must also be noted that female gorillas and chimps will travel to other troops at great personal risk in order to mate with males who are not their close relatives.

Genetic sexual attraction (GSA) is a well-described phenomenon in which siblings (who have similar genetic makeup and thus MHC profiles) will often become sexually attracted to each other if separated at birth and later reunited without being aware of their relationship.

This tendency is usually moderated by the Westermarck effect, in which people who grow up together from an early age (even if not related) will preferentially seek out others as mates. Anecdotally, the Israeli kibbutz system promoted a Westermarck effect in which (unrelated) young men and women who worked together on the farms for a long period showed no romantic interest in each other.

These results are interesting, and seem to make sense from an evolutionary standpoint, but we must always be careful with conclusions that--rightly or wrongly--suggest that what we have already decided is true.

Why hesitate to synthesize when these results make perfect sense? Sexual attraction appears not to be constrained by perceived relatedness when uncoupled from fertilization. At the same time, heterozygosity is favored in fertilization couplings (apart from sensual, loving, or intimate), reflecting an overall mating system (featuring male competition, sexual selection, etc.), and sexual reproduction itself, which promote not just heterozygosity but phenotypic variability as raw material for selective change. On a broad phylogenetic scale, heterozygosity becomes a matter of group survival only to the extent that groups are relatively unfit, needing to generate variability for change. Lots of lines do pretty well asexually.

That is to say, humans are driven to produce offspring slightly different than themselves each generation, because the line including humans, as an ongoing experiment in fitting with its environment, has not been doing so well.

Genetic and mating system mechanisms promoting variability and change would become increasingly important in populations unfit in their environments. We become puzzled and attempt to provide explanations in terms of positive fitness and adaptive value, rather than lack of fitness and drive for change, only because of our cherished and absurd view that Homo sapiens, as species go, is relatively fit in its environment.

Take a look around.

Add to that the fact that fertilization between people with similar MHC profiles often leads to miscarriage (at a higher rate than expected, that is), and I find your synthesis very interesting.