I understand that they base their research on CVE data because it offers normalized quantifiers of severity and scope, but in my experience vendors by and large don't bother with CVE's for API bugs even when the affected primitive is clearly malfunctioning (memory or correctness issues).

I've been deeply fuzzing cryptographic libraries for a few years and found about 130 bugs [1]. The vast majority of these did not receive a CVE. Now some of these are merely theoretical, others will only manifest under particular circumstances like specific calling sequences, others were caught in the development phase before landing in stable releases, but a number of them are outright vulnerabilities. The usefulness of CVE incidence is questionable when it is so strongly influenced by the vendor's propensity for reporting these.

[1] https://github.com/guidovranken/cryptofuzz#bugs-found-by-cry...