As always, thanks for jumping in!
We use SyncroSoft's PDF Chemistry for PDF publishing. And it behaves correctly - it doesn't create navigation hierarchy from submap titles. My previous comment was that topichead levels added inside a <bookmap> would alter the book PDF's TOC, so modifying the <bookmap> is not an option.
However, your mention of <navtitle> within <topicmeta> pointed me in the right direction! I get the desired result by variable-izing the book titles, then creating a top-level map as follows:
<mapref href="book1.ditamap" format="ditamap" keyscope="book1"/>
<mapref href="book2.ditamap" format="ditamap" keyscope="book2"/>
When I publish this map (either with the DITA-OT's html5 transformation or with Oxygen's WebHelp Responsive transformation), each book has its own hierarchy item, named after the book title. Success!
With this success under my belt, I started poking around in the preprocess2 pipeline to see if I could automate the creation of these <topichead>/<navtitle>/<mapref> constructs with a user-specified attribute as a convenience:
<mapref href="book1.ditamap" format="ditamap" keyscope="book1" outputclass="topichead"/>
<mapref href="book2.ditamap" format="ditamap" keyscope="book2" outputclass="topichead"/>
It is achievable, but there are some complexities with ensuring that book titles with profiling conditions and book-local variables are handled properly. I might keep at it just to learn more about the processing pipeline. But for production, I think it's better to stick with pure DITA without adding yet another plugin dependency.