Documents and humanity in general are complicated and various. You're never going to be able to fully encapsulate the complexities of a document into any (structured) DSL as a result. The recommended approach, failing all other options – even to the semantic people – is to use some standby element, such as DIV or SECTION. Unless you're including people who would say "just change your document to not do anything not representable by Semantic HTML". Add to that, that the list of supported elements aren't even close to approximating what you'd write in 50% of documents.
Because nested layers of DIV are apparently "semantic"