Another possible gotcha with the data in this graph: the "not selected" number appears to include pages that used to exist but which now 301 redirect to another page.
I know this because one of my sites with 40,000 pages had 2 pages about each topic. I combined into one page about each topic and 301 redirected so that I have 20,000 pages on the site. As Googlebot found these redirects over the course of a few months, the "indexed" line fell from 40,000 to 20,000. However, the "not selected" line grew from 1,000 to 20,000 over the same period. A mirror image of the "indexed" line.
I wouldn't expect 301 redirects to be considered pages that were "not selected" for the index. I would expect "not selected" to be actual pages.