One thing I think is often overlooked in comparing brains with computers is that while they may be ahead on efficiency and density, computers win on many other metrics eg. They are easy to reprogram, they don't need sleep, etc.
And I'm curious how exactly brains are measured for operations per unit <blank>
Estimations via a variety of methodologies that all yield different answers. Since brains and computers run such fundamentally different "programs", the details matter less than you might think, because the whole question starts as pretty hopeless in the first place. One approach is "what would it take a modern computer to fully simulate a brain", which is in some sense throwing a huge advantage to the biological brain because the same measure in the opposite direction makes biological brains look really bad, far worse than this measure makes computers look!
Still, while the computations may not be generic, there is certainly a lot of stuff going on in a biological brain, no matter how you look at it. Note the graph is on a log-log scale; differences of opinions will tend to move the biological systems a little bit one way or the other on the graph, it won't tend to suddenly fling it halfway across.