We've seen this type of stat many times, and it's misleading as . What they mean is that if Kawhi is still guarding Harden by the time he shoots, he doesn't shoot well. That makes sense, because it means that Harden hasn't gotten around him, forced a switch or something else like that. That inherently biases the sample to good Kawhi possessions. Kawhi's been the best Harden defender so far, don't get me wrong. But he pretty much went up and down with him last game with James scoring at nice clip. Trying to say, "It was actually on Pau when Harden beat Kawhi and got the layup" doesn't actually fix anything.
The Spurs are already playing Kawhi on Harden. The solution to checking him isn't more Kawhi, as the stat would imply. It's trying to shore up those possession that end in "others" or by going back to the Game Three strategy of not caring if Harden scores.