1. Yes, the filters don't work, except to keep certain transactions off your local mempool.
2. Most users probably don't care. The OP_RETURN limit was an ineffective tool to discourage data storage. Hence its removal in 30.
Discussion
1. Then why, to your original point, do “enough people need to run core to keep people from censoring transactions “?
This is not logical.
My point is the filters WOULD have some effect if virtually all nodes, including miners, enforced them. So when I said "enough people need to run core," I meant just a few will do, and those transactions will still get mined.
But then why does the filter “for my own local mempool” have any effect on mining? Even if it’s 100% of the nodes? How is that possible? This is a logical dead end for you.
Miners have mempools. If they are filtering, and building block templates from filtered transactions, filters work. See recent Ocean blocks.
So if 100% of the non-miner nodes use knots, and all the miners use core then there’s no censorship concerns, right?
So why are the core apologists berating me for my choice of software? They should only be convincing the miners to run core, since that’s all that matters