There is now a more-or-less final Microdata to RDF algorithm published which is different than the one in the current, official HTML5 draft  (that Ian Hickson has publicly revoked). However, Any23s extractor uses the old scheme according to a comment in its source code, which refers to . However, this is exactly the algorithm that Ian Hickson rescinded at some point. Unfortunately, the official working drafts have not been updated for a very long time, but if you look at the editor's draft , you will see that that section has been entirely removed. Instead, there was a Semantic Web Interest group task force that discussed the issues, and  is the result of this discussion. It would be nice if this would be reflected in Any23 in the future.
[Condensed from an E-Mail conversation with Ivan Herman]