The infamous "screen shots" from 2004 demonstrate this -- and New Hampshire is a good place to look, since there was a double-digit exit poll discrepancy
and a large partial recount that supported the original results.
http://www.exitpollz.org/exitpolls/NHam1224CNNScreen0035.pdfshows men 52/47 Kerry, women 58/41 Kerry
http://www.cnn.com/ELECTION/2004/pages/results/states/NH/P/00/index.htmlshows men 47/52, women 54/45
(the discrepancy was larger than the first tab indicates, because the initial projections incorporate pre-election expectations)
You're of course absolutely right that they have no way of knowing whether the original estimate was "off" for men, women, or both, in what proportions. They don't adjust the tabulations individually; they adjust the weights for each respondent. In this case, that should mean that the estimates for men and for women change about equally (which is true, although given rounding error, we can't tell exactly how true it is).
Insisting that we know the original exit poll was right is the wrong way to go, and will blow up in your face -- if not this time, then next time the exit poll is wrong.