For your point 3, even xml hierarchy does not provide any semantics including that there is necessarily some hierarchy implied by XML containment. Not even generic ones like connects-to, has-a, is-a. It is no different to IDREF or a URL. XML is just a transfer syntax to go from common bytes to some useful graph-structures. (And back?)
Consider a document format that has page-break marker elements in otherwise semantic but sequential markup: the pagebreak attaches more to a position in the data content of an element than being in any 'semantic' relationship to the element. (And,yes, better as a PI probably, in the abstract.)
Regards
Rick