Also, even with identical DNA, that doesn't mean the twins will be exactly identical: if things like fingerprints are only very similar instead of exactly matching, then there's reason to think that the fine brain structure can vary as well -- perhaps enough to straddle sides of a disorder.

So in addition to twin frequency, it would be worth analyzing twin spread: what is the variation in the behavior of twins (in the background), and is the spread in autistic/non-autistic twin pairs statistically similar, ie, are we just selecting for normal, expected variation that happens to fall close to a boundary?