The point I was making was not that tech support is likely to have an "unusual problem bias," but that being correctly calibrated with respect to usual and unusual problems will tend to appear like a "usual problem bias" when you examine in isolation the cases where they're wrong, because you would tend to observe cases where they need a significant amount of evidence to persuade them of the presence of an unusual problem, but not an unusual one.

If you examine cases where they're right, you may find a large number of cases where the customer insists that the problem is not addressed by the tech support's script, only to be proven wrong; these often appear in the horror stories posted by tech support. Thus, tech support may to some extent be rationally discounting evidence favoring unusual problems.