And he hasn't advocated using something other than the expected case when evaluating a non-transparent charity. What you may infer is that he believes that the worst case does not significantly differ from the expected case

That's not what Multi said. He said we should assume the worst. You only need to assume something when you know that belief would be useful even though you don't believe it. So he clearly doesn't believe the worst (or if he does, he hasn't said so).

I don't use the word "assume" in the way that you describe, and I would be surprised if multi were.

He also said that he "believe[s] that reducing existential risk is ultimately more important than developing world aid." How do you go from there to supporting StopTB over SIAI, unless you believe the worst?

Here I think we more-or-less agree. On my reading, multi is saying that, right now, the probability that SIAI is a money pit is high enough to outweigh the good that SIAI would do if it weren't a money pit, relative to a tuberculosis charity. But multi is also saying that this probability assignment is unstable, so that some reasonable amount of evidence would lead him to radically reassign his probabilities.