I found the fact the CX cannot adapt <score> tags (T139457) by coincidence.
There might be more features in MediaWiki and its extensions that CX cannot adapt and of which we aren't aware yet. If possible, we should log such things, to make discovering them less random.
This seems relevant to identify how well the tool supports different types of content (T191389) and references (T200786). We could better identify the most problematic elements and capture the progress (i.e., the improvements in support should reflect on the decrease of issues).
@santhosh is this something that would be useful to get a better idea of the adaptation process for templates (and references)? Is it worth capturing the approach used for adaptation in those cases (along the lines of what was manually done in T224721#5258865)?