Thank you Stephen, that is helpful context, especially the differentiation between Genre and Subject headings. (although I think the distinction might be confusing for casual readers....). But in context of the data available, that makes sense and is helpful.
My use case is around book recommendations and book browsing, so the "bookstore" analogy of Bisac is probably on point.... (I have a personalized book recommendation engine, The Hawaii Project (www.thehawaiiproject.com)) - the recommendations are based on metadata available from amazon, one's reading history, and what book reviewers are writing about). So both fiction and non-fiction are in scope. But the genre/topics I provide browsing for are limited, and I'd like to be able to go deeper (i.e. one can ask for recommendations about "mysteries" at present but not specifically "historical mysteries", nor browse that way. (Search is a different matter, and easier to handle).
The harvard data is open and incredibly rich, so am looking for the best way to leverage it....vs licensing fairly expensive commercial metadata.