His math and reasoning is right, but what he does wrong is treat the water in the water block as an isolated system.

In fact, he should be treating the whole water loop as the 'water reservoir' that is undergoing total equilibrium with, say, the CPU block. And if flow is high and therefore you're getting more water through the rads per unit time, and therefore you're keep the whole water loop cool(er), then your delta of the block to the water will be smaller than if you allowed low flow and had all the water equilibrate to a higher temperature.

I believe that's the flaw in his reasoning.