Not all studies use the same definition of sexual assault. Surveys in particular are subject to question wording and question order effects. As army1987 notes, the ~20% proportion is for sexual assault, not just rape.

Keep in mind that the object-level question here is whether a rape-reporting rate of 12% can possibly be consistent with a ~20% sexual assault rate. Will (and the media in general) misstated the class of events to which the "12%" referred; Will then stated that it could not possibly be the case that the 12% and the 20% were consistent. This is a very strong claim, which means that checking/refuting it is easy in absolute terms. To refute the argument, it is not necessary to have precise estimates -- it is only necessary to show that the statistics being reported are broadly consistent.