Dot, I agree with Roberto!
As part of my presentation at Kalamazoo last year I described how (for Electronic Boethius) I modified some of the manuscript description elements for use in transcription (for example, marking rubrics in the text using the <rubric> element). This obviously wouldn't work if one wanted to both describe and transcribe a text in the same document (as most probably would?) but for EBo I think it worked okay.
The only reason it wouldn't work right now is because the <rubric> element is not a member of the class of elements which can appear within transribed text. But absolutely no reason why it couldn't be added to an existing class or a new one (model.titleLike maybe) in a P5 application, or better in a set of proposals for changes to the class system...
I personally would like to see an effort to bring manuscript description and transcription together more closely. I think that this would be done best not from within the TEI (and I speak as a member on the TEI council), because the interests of the TEI are quite general.
I don't understand either of these assertions. The TEI has lots of very specialised tagsets and aims to support very specialist usages within a general framework. And changes to it have to come from within its user community, of which after all DM is a major constituent.