If people make biased (http://en.wikipedia.org/wiki/Bias_of_an_estimator) guesses you will still have a problem. I think that's often what's meant by "rational": that people, in aggregate, make unbiased guesses.

The truth is not so - people get irrationally exuberant or are afraid to cut their losses, etc. These are problems of statistical bias of their estimations - and the opposite is assumed in many (most?) economic models.