Could you maybe clarify what happens to entities when they are included in the content? Several RSS readers I experimented it needed double encoding to show correctly (i.e. < although I use namespaces and XHTML.)
If you avoid type="html" you should not need double escaping in Atom. SImple escaping should do. If you do use type="html" you will need double escaping to represent characters such as less-than and ampersand.
I recommend avoiding type="html" if possible, so if you can follow this advice you should be able to avoid escaping puzzles.