My thoughts on this are that if you are making an EPUB book from an original printed version then what you see on the original page is either italic or bold, not the concept of strong or emphasis. You might guess why the author chose one form or typeface and you might be right.
On the other hand if you are starting from a clean sheet and are the author of the book then you are in a position to know why you want something in italics, bold or even underlined and can use the semantic approach.
The idea of disconnecting the meaning from the presentation is a good one - it allows for an ebook to be read by say a text-to-speech engine or a braille reader (though this might not be the best example after looking at
http://www.brailleauthority.org/form...formats05.html ) but think about it - some readers will render <em> in bold, others in italic; if you want the names of say ships to show up in italics the only sure way is to use <i> (or use css font-style) that way you either get italic or no formatting if the reader doesn't support it.
BobC