I think that we might have different definitions of how good or bad EndNote is. As an individual neuroscience researcher who has his entire PDF library referenced in EndNote it does work well enough, not that I like it. You seem to be approaching it from a more technical standpoint of usability across many disciplines and citation types. There EndNote is indeed trash if you routinely require citations not found in EndNote's templates.

As for the technical rigor of the programming, 99% of the citations I use in my manuscripts are books and journal articles. There might be a few conference presentations, personal communications, and whatnot in there as well, but that is about it. This reduces the scope of the problem to something that IS relatively easy to implement software-wise. The goal of CSL seems to be the creation of a catch-all language that can represent any citation. This is a far different problem in terms of difficulty.