GiveWell spends a lot of time making estimating their performance easier (nearly everything possible is transparent, "mistakes" tab prominently displayed on the website, etc.). And I know some people take their raw material (conversations, etc.) and come to fairly different conclusions based on different values. GiveWell also solicits external reviews.

This is nice to hear. Still, you have to trust them to report their own shortcomings accurately. And if more and more people join EA for status reasons, GiveWell and related organizations may become less incentivized to achieve high performance.

Everything else you said, I agree with. Are those your totality of reasons for not endorsing EA? If not, I'd like to hear your others (by PM if you like).

Mostly these are the reasons I can think of. Maybe I could also add that donations to people in impoverished communities might create market distortions with difficult to asses results, but I suppose that this could be lumped in the estimation difficulties category of objections.