Are these debates just backwash from the introduction of HTML as a rough-and-ready, who-cares-about-strict-semantics markup language, and the subsequent stumbling attempts to separate structure and presentation with XHTML and CSS?
To me, even though I've never used a screen reader, knowing they exist and that it's important as a web designer to acknowledge their existence has been crucial in recognising, for example, the difference between EM and I. The question "What is 'emphasis' supposed to mean?" seems odd in this light. It means emphasis ;-) I is just an overhang from HTML's mix of presentation and structure. The visual, printed convention for, say, book titles, is italics. But you wouldn't normally want such a title *emphasised* if the text was spoken aloud. (Theoretically, in this case you should extend XHTML with a BOOK element, or use a CLASS on a SPAN and apply the italics with CSS. I usually stick with I for backwards-compatibility.)
Some words you want to emphasise. Visually, that usually translates as italics; aurally, it means intonation and volume. The web *is* primarily a visual medium, at the moment. But it seems wilfully ignorant to disregard the efforts to separate structure and presentation over the last 6 years or so.
In terms of writing web content, as both a writer and web designer I've got used to writing with markup, and it's always my sub-vocal expressions that guide the markup I use.