One type of response to this kind of argument is that it doesn’t strive to provide a model for that actual experiment. This response is entirely irrelevant, though it is understandable how it would appeal to physicists. If you write a simulation, for instance, it is necessary not only that the simulation disprove the argument, but that it look as much like the published experiments as possible. Thus we have papers like this one: http://arxiv.org/abs/quant-ph/0105034
(An error in application of Bayes’ rule is just one of the ways the problem can be characterized, and happens to be the first on record and probably the one most accessible to non-specialists. Lots of people know probability theory, and some of them actually trust it to give the right answers if you strictly follow its rules.)
