I think one common error people make is to say "Oh, well, rational actors are wrong therefore I can throw rational actors out of the model entirely, everything goes, nobody knows anything, case closed." As you say, it's still a good first approximation, and to the extent various irrationalities are found, they too are actually reasonably regular and predictable as well. Irrational isn't a synonym for random.

(Moreover, there's a certain relativity to "irrational"; as tempting as it may be, irrational isn't a synonym for "disagrees with the given economic theorist" either....)