We're trying to avoid hi because it's so widely used in so many different ways. We would also prefer that all styling be applied for semantic reasons, so rather than ‘italics’, think ‘journal title’ or ‘emphasis’ or ‘foreign word’. Also, what you believe should be in italics might in our style guide be rendered in quotation marks, or not styled at all; that's why it's easier if you identify things and let the system style them.
Literal quotation marks can be straight double, straight single, curly double, curly single, or (if you happen to be on a non-English keyboard) a range of other symbols. All copyeditors are familiar with the tedium of checking that they're all the right form, and that all the inital ones are opening ones and the closing ones are the matching closing ones. It's much simpler if you tag your text as quote, soCalled or whatever, and let the XSLT provide the quotation marks in a reliable way.
The TEI is a huge standard; there are lots of available approaches to encoding any given phenomenon, and every TEI user has their own habits and preferences, arising out of their history and the projects they've worked on. If we accept submissions in any valid TEI (‘tei_all’), we inevitably spend many hours re-encoding them to get something that will work with our system. It's much simpler if we let the schema do the work for us.