IIUC, I think the distinction between sourcing and assessing is not relevant here. Assuming the females in the pool of viable candidates have already been exhausted, then sourcing only females is suboptimal, as they are less likely to pass assessment because (by construction) they are not in the set of viable candidates.
It will always make the most sense to source the best people without regard to anything but job skill. If the best people aren't diverse, then chasing diversity requires you to look outside the set of best people, by construction. You can't have your cake and eat it, too.
