You hit the nail on the head, here. There's a fine line between "bold, fearless, and usefully foolish" and "too ignorant of an extremely complicated domain to accomplish anything useful", and in medical informatics that line is even finer than in many other fields.
> I don't think any company in their right mind would use Metathesaurus as their base vocabulary, may be synonyms, mappings etc but not as a coding vocabulary.
Just to chime in (in case anybody reading this comment thread is thinking about getting into this stuff), you're 100% correct: the UMLS is in no way supposed to be used as a coding vocabulary. If you try and use it as such, you're Doing It Wrong(tm). That's what clinical terminologies e.g. SNOMED-CT are for.
Of course, that doesn't mean that plenty of people haven't tried- it's just such a tempting-looking resource...
That said, I certainly wouldn't want to dissuade anybody from learning how to use the UMLS, or from using it for its intended purposes. It's a really valuable (and easily accessible) vocabulary resource.