I dont think they're lying.

But these are ensemble averages, and presumably bad testers were as common before as they are today.

Or maybe not. Maybe the problem is the composition of the ensemble. Seems important to figure this out.