I think this is a good approach. If you are using <text> element then what you are privileging in your markup are the intellectual structures you find in the text. Imagine a a parallel example: A heading of a division of a magazine article that is provided as a cut-in from the right margin into the middle of the second paragraph. It would be clear from context that this was a heading for the section or article and in coming to that interpretation I would mark it as a <head> immediately inside the <div> at the top. I might indicate through @rend or @rendition an informally or formally (respectively) description of where this is ('right-margin-cut-in-header' or the formal CSS to float it there). I think the same is true of a post-script. Moreover, I don't think indicating the place by an anchor is necessary or accurate if you are modelling your semantic understanding of the text.
If, however, you do care precisely where this is, then I would be referencing a <zone> for the <surface> either in a <facsimile> (if you just care about the location either on an image of the surface or its place in a coordinate space you've created) or a <zone> in a <surface> in a <sourceDoc> if you wish to reference a non-interpretative transcription from your interpretative one. Since I don't tend to want to do that I would either do the @rend/@rendition or <facsimile> methods personally.
Dr James Cummings, [log in to unmask]
School of English Literature, Language, and Linguistics, Newcastle University