The arithmetic works, if the underlying stats and assumptions are true - a big if, of course. As follows:
If men commit 90% of assaults (is that all??), women 10%, this means a man is 9 times more likely than a woman to commit an assault. Say we have 1000 users of whom 995 are women and 5 men (i.e. 0.5% of users are male). Say each woman commits x assaults, so there are 995 woman-caused assaults, then each man commits 9x assaults, so there are 45 man-caused assaults. So in total there are 1040 assaults of which 995 are woman caused: 995/1040 is indeed 96% (rounding to nearest percent).