That table is unfortunately quite old. I can't personally say what have changed, but it is hard to put much confidence in the relevance of the information.
Yeah, also it doesn't compare actual implementations, just plain checkboxes. I'm aware of two specific substantial performance regressions for musl: exact floating point printing (it uses Dragon4 but implemented it way slower than it could have been) and memory allocator (for a long time it didn't any sort of arena like pretty much every modern allocator---now it does with mallocng though).