>t would be reasonable to say that CAPM is a mathematical result, and the evidence suggests the mathematical model required for CAPM simply fails to correspond to reality. But many finance folks, especially in academia, won't admit it.

In other words, "Things are more complicated in real markets". I don't see any problem with wanting descriptive theories. Economists are interested in building models whose parameters can be tweaked in ways that correspond in a clear meaningful way to a possible actions (e.g. change government expenditure on X by Y). If you try to just maximize predictive power you're missing the point. Economics can be used to predict the future, but I would say the main goal of the field is to produce theory that will provide valid methods of altering the future.