The counter argument is that in practice, the true fitness function is only very roughly approximated by the selection heuristics used. And if - consciously or otherwise - $ismale happens to be weighted positively in the selection heuristic (more so than any positive effect maleness might actually have on their performance in the job), a quota which artificially inflates the number of females might actually go some way towards redressing that mismatch.

Obviously this is considerably more likely to actually be the case in some businesses and industries than others.