The only way I can see this discussion being fruitful would be to actually build a Bayesian network for each top-level claim and then individually assign probabilities; this at least would be an interesting exercise.
An interesting idea, but in my opinion, the only way to build an accurate Bayesian network for the top-level claims would be to examine the context whence those claims come.
You could have one Bayesian network to show how likely Mormonism is, and a separate one for how likely it is for a holy book to say that otherwise. Alternately, you could just give the ratio of how likely it is if Mormonism is true and how likely it is to be stated otherwise. I suppose you'd have to calculate for a given religion, and then multiply by the number of religions.
In this comment thread, I gave the following idea, on the topic of a method by which one might judge the Book of Mormon from a rationalist perspective:
I would appreciate feedback on this idea, for an admittedly selfish reason: I am trying to instigate in myself a Crisis of Faith. So, here are the questions I pose to you: