Seems much steeper than the original recs... I ran 2016 on something not much better than the minimum here maxed out at 60-90 fps.
So it's either way better looking, or worse optimised, or the test matrix was limited by budgetary/time/space concerns? Which isn't bad. Maybe focusing on more modern stuff makes more people more happy. Might even work fine on the old shit after some patches. But no test lab is going to have 80 different configuratuons of a 2016 PC I guess.