I don't know this for a fact, but others have looked at the Amazon rankings and evidently it doesn't take a large change in volume (number of books) to drastically affect the rankings. So that big drop in the rankings isn't such a large drop in actual books sold.
I do look at the estimated sales from the rank, and in some ways you're right. By this one calculator I use, any given rank gives you a certain sales per day and sales per month (not the same since Amazon rank weighs daily sales morr than monthly). And you're right that the absolute number isn't that big. But then think of the percentage. Until relatively recently, the 6000s were the bottom ranks for P2, now it's the 10000s. That is a 35-50% drop in Amazon sales in a couple of months. Is that trend a good one if this sticks for long? That's what I care about. Because of all the qualifiers and uncertainties, the absolute number is not as important in trying to determine the health of the game. If things stay where they are or continue to decline like they have ever since August, I'm going to doubt rosy prognostications without some other indicator of sales numbers or general interest.
People like to argue that these sales don't mean as much because of the direct and store sales. However, they are taking on new meaning because of recent events. Diamond Comics, which I think is a major Paizo distributor as far as I can tell (correct me if I'm wrong), is no longer shipping new product to comic stores because of the coronavirus. We may have a very unfortunate test of how good Paizo direct sales are if other major distributors follow.