There is an overwhelming amount of evidence that CAPM simply doesn't work. Beta is not a good description of risk. No wonder analysts have such trouble forecasting stock prices when they routinely use beta as a key input.
CAPM woefully underpredicts the returns to low beta stocks and massively overestimates the returns to high beta stocks. Sadly our industry seems to have a bad habit of accepting a theory as a fact. This is at odds with a scientific approach which likes to test theoretical models by subjecting them to empirical evaluation.
The CAPM fails because its assumptions are clearly at odds with reality. Two of the critical assumptions in particular stand out. Firstly, that we can all take any position we please (long or short) in any stock with absolutely no price impact. Secondly, that everybody uses Markowitz Optimization (MO) to assign portfolios. Even Harry Markowitz himself doesn't use MO! The CAPM is, in actual fact, Completely Redundant Asset Pricing (CRAP).
Instead professional fund managers seemed obsessed with tracking error. To a tracking error ...