I've never dived (dove? doven?) into the output of an HTML minimizer, but I assumed this was one of the things it would do - remove all optional tags while retaining the same DOM tree output.
HTML was intended at first to be hand crafted by scientists. That it should be forgiving is a good thing, because who wants a "this page has imperfect html" message.
Useful for code-golfing HTML, but the complexity of rules sounds like a maintenance nightmare for professional use.