Of course, that's from his last update before the election. It'd be more interesting to look at his projections from many months out, when not so many of the races are going to be in the 90-100% range.
Alas... more were. At least for House races. (There was a formula change in there somewhere, too, but it's minor - led to a lot of races being downgraded from 99 to 96 and stuff like that.) Logically, no House race should be rated 99% months before the elections and then decline to 70ish or 80ish once it becomes clear that it's competitive(ish) after all. Yet that happened a lot in 2010.