Isn't that why we have the smart/rational distinction? One way you might see it: smart = generates relevant logical information at a high rate, rational = processes this information in the right way so as to come to true beliefs. (This is vague but I hope you can see the intuition.) Aumann and hypothetical-Perelman both seem able to generate interesting pieces of reasoning better than almost all people, but seem to sometimes have trouble fully stitching together and accepting the implications of the interesting true ones when not disciplined by standards of mathematical proof.