It's increasingly acknowledged, even in the "soft" sciences, that r-squared is not an appropriate analysis tool.

Anyway, wasn't the IQ test originally created to determine whether or not someone was mentally disabled? It seems like with more modern statistical methods it still could be useful measurement for that purpose, even if it doesn't accurately "scale" beyond 80 points. Ultimately, if you just want to answer the question "Does this person have significant cognitive deficits?" it still has its place even if it's not a good measure for broader intelligence.