This is an excellent analysis, but of course what's also needed is the date at which an article was made OA, relative to the date it was accepted for publication. Research progress depends upon immediate access, yet some publishers embargo author self-archiving by 6-12 months or more. (63% of journals, including almost all the top journals, endorse immediate OA self-archiving by their authors; 37% do not.)
Timing data are difficult to collect, but future studies may be able to make estimates of this too. (Even without the timing data, however, the results from Professor Björk's group are very timely and useful.)