And if you are referring to in-Spark transactions, the same issue exists in an in-LSP transaction... I.e. the sender and receiver are connected to the same LSP.

I recommended you a few times in the past not to overhype the privacy state of Lightning. I suggest you align your documentation with this:

Lightning Privacy Research

Lightning Privacy Introduction | Lightning Privacy Research

An introduction to Lightning privacy

You are likely right about Phoenix, it's true I haven't looked into that implementation. "If the LSP knows the destination node, it knows the destination." Yes, but in the case of all the standards compliant LSPs, we DON'T know the destination. We only know the hop before the LSP and the hop after the LSP. Only the sender knows the actually destination. "the LSP effectively controls the path because the graph information is fed by the LSP". This is an extreme edge case of a malicious LSP, and to pull this off at scale would be extremely complicated.... and also, the LSP could easily get caught by a savvy Lightning engineer who was troubleshooting payment problems.