It does seem like survivorship bias could be a factor. The article talks about how some of the old models were off until they "corrected" them. All right, and did they apply the same scrutiny to the apparently "correct" ones? "All of the models predicted this (except for the ones that didn't.)