Compression isn't the only mechanism by which you can lose information in the process of going from RAW to something standardized and hardware-neutral. So simply using lossless compression doesn't actually mean you're preserving everything. If you really want lossless compression, you probably also want to preserve information about how the raw image was transformed during demosaicing, denoising, white balancing, etc., all of which are somewhat subjective instead of being exact 1-1 transformations where there's a single right way to do it.
All those algorithms the parent is talking about are custom made per sensor generally. The ideas are the same, but they have to be tuned for each sensor and as the parent said, it's incredibly subjective and every company / vendor approaches the problem differently depending on use case.