That does sound like Problems with COG images downloaded from the Planetary Computer API, though in that case I thought it was only related to COGs and not the JP2000 files, but I might be wrong about that.
I reported that upstream in What is the valid range for the TCI product? - Optical Toolbox - STEP Forum, and ESA (who is at least involved in sen2cor’s development) confirmed the bug. I don’t see a confirmation of a fix in that post but I thought they’d announced a fix.
Assuming it’s fixed in sen2cor, you’d need to check with whoever is hosting the L2A imagery (is that Synergise?) and see how they’re getting it. I thought that Synergise was getting L2A directly, not not running sen2cor, but again I could be misremembering.