One of the best conclusions i have seen in response to P/E arguements, is that P/E is one dimensional. I.e. P/E in 70s may have different bond/interest/inflation conditions than 2000s, although both look the same.
When shiller makes the P/E argument, I don't see him taking into account the premium over bonds during the periods evaluated (although, he does look at real returns).
Also, for individual stocks; two stocks may have identical trailing P/E, but ignore things like earnings growth projected out into the future.