This is excellent and exhaustive. Assuming you'll be submitting this for formal publication, here are a few suggestions:
–Coverage of book chapters (14%) sounds puzzlingly low, especially given that 82% of requests were successful.
Perhaps requests for chapters are redirected to the entire book, which is then indexed under its own DOI? This might be a strategy to minimize the number of downloads needed from publishers. Delivery of the book might then be logged as a successful request for a chapter DOI, but it would not be added to the main index. A modified search to check this would be simple for some publishers, e.g. Springer, who just append an underscore and chapter number to the book DOI.
-In the abstract, the phrase "all 81.6 million scholarly articles" would be more in keeping with the main text if hedged (e.g. "Crossref-indexed" or "with DOIs").
The case for using crossref could perhaps be strengthened by comparing overlap in both directions, e.g. "for recent years, close to 90% of articles in Pubmed, WoS and Scopus have DOIs, but only y% of articles with DOIs are indexed in Scopus, the largest of these. Thus, crossref is by far the largest database of the formal scholarly literature, approaching complete coverage for recent papers."
–rather than "closed access", which could be taken to include e.g. embargoed and private items in repositories, I think Suber's term "toll access" would probably be understood by a wider audience.
–The argument "the scope of this resource suggests the subscription publishing model is becoming unsustainable" could perhaps be fleshed out a bit, noting that the recent cancellations of all Elsevier subscriptions by major universities are unprecedented, following decades of empty talk about the serials crisis.