It's easy to think of a few "not all bugs are created equal" reasons this might be misleading.

Weighted averages of CVSS by OS might have been more useful, for example, and dis-aggregating Microsoft OS versions but not the various OSX or Linux distros/kernels (and lack of Android) also paints a potentially misleading picture.

With any exercise of this type, the conclusions one draws are typically encoded a priori in the way the analysis is framed and defined.