The multi-document processing, you need, is in our TODO list, but it is not implemented yet.

The only solution for the time being is to preprocess the HTML before it goes to PD4ML and to remove all and tags. Unfortunately all CSS styles defined in the documents are going to merge, which can result.