The problem is that the extended norms are even less calibrated than the rest. I know a tester who refused to use extended norms because the child had 2 ceiling scores (18 or 19) in separate sections of the test (performance/verbal).

I don't know how far the scores can extend; but I have heard of numbers in the mid-twenties. This would certainly go beyond the SD for that; but how realistic they are?

I agree with what you say and testing. I was also referring to Ruf's levels; so many of her indicators are academically related: child knowing letters, counting, etc. If I went strictly by her early development, Ruf's levels would have seriously underestimated her current achievement levels. There are many kids like her who do not fit in that model, namely many 2E children.