I'd be inclined to disagree with part 3 as well. If you are biased, your probability estimates are off. The fact that everyone's probability estimates are off to at least some extent does not change the fact that yours are off.

I don't think Slifkin is talking there about wrong or right in the Bayesian sense. To translate that into a more Bayesian form it might be something like "Just because one is biased doesn't mean that one's probability estimate is drastically off" or something like that.