They are probably just using 'men' and 'women' as an approximation that sort of matches the distribution of human beings. Without diving into the exact individual's relationships with sex/gender, you could get a somewhat accurate approximation of how a society operates. Then, from this 1st order approximation, if one was inclined, they could refine and add to the results by further breaking down the input parameters. I don't see any issues with approximating a binary distribution of male and female, even though everyone knows this is not a perfect description of reality. The fact that there is deviations from a binary is good, and can always be tapped into to add alternative views and refinements to the results.