Hmm this is interesting because I'm actually not sure whether to treat the survey numbers or the funding numbers as the better signal of what the community actually believes! Surveys capture what people say when asked in the abstract. Funding decisions capture what people do when they have to make tradeoffs with real stakes. To me, neither is obviously more "true"
When I answer a survey about ideal resource allocation, I'm not doing the same kind of thinking as when I'm actually deciding where my money goes. surveys feels more like "what do I want the world to look like" whereas the donation feels more like "given my uncertainty about sentience, tractability and what others are funding, where's my/our marginal money best spent." These can reasonably diverge as you can see.
That said, I do think your lost in translation hypothesis is true in a way. The discourse is more AI-dominated than the MCF survey would predict or at least feels so. And if the community's stated preferences are roughly right but people are deferring to a perceived consensus that doesn't actually exist, that's a coordination failure worth pointing out!
Also resonate with @SiobhanBall's point about psychological friction. Animal advocacy is the one cause area where most of us are personally implicated and that creates a kind of discomfort that's easy to avoid by just... focusing elsewhere. A kind of moral responsibility offsetting. I'd be curious whether the stated/revealed gap is smaller for vegans in the community!

Yeah well a lot of neglectedness discourse does end up being "here's a big number, here's a small number, therefore neglected" without asking whether those numbers actually inform anyone's marginal decision? I think shrimp welfare is also interesting because the bottleneck isn't "people who care about shrimp don't know it's important" it's that very few people care about shrimp at all.

One thing I'd find useful is more on how to handle cases where something is at the edge of your circle because of empirical uncertainty rather than moral uncertainty. I might be unsure whether to include shrimp because I'm uncertain about their sentience (empirical), or because I'm uncertain how to weigh their interests even if sentient (moral). These feel like they should cash out differently. The first suggests investing in research, the latte4 suggests the "neglected because few include them" logic you allude to.