Firstly the data aren't usable for this kind of analysis even if you bucket them differently: https://www.cvedetails.com/how-does-it-work.php
Secondly, we know nothing about exposure, because we don't know how the relative levels of attention or how quickly the vulnerabilities are patched.
A high discovery rate is as indicative of the attention given to the platform as it is to the underlying number of vulnerabilities. A high discovery rate plus relatively high rate of distribution of patches could means a system is less vulnerable.
I'm not saying this is the case - just that the data presented are simply useless for ranking software vulnerability levels.
If, for example, researchers found 10k vulnerabilities in windows 3.1, it wouldn't make sense to ding microsoft for an insecure operating system, because no one runs that and it's not reasonable to expect support.
I would not be overly surprised if it turned out somebody was still using Windows 3.1 - I know of at least one company still running NT 4 on at least a few select machines. (This does not invalidate your point, though.)
And if you even ignore type of product, then Adobe goes first, but this has no impact on "which OS was the most vulnerable".