But the OP's points have more complexity than they think (in my main comment[0] I mention depth being missing). Let's take the quick math one for example. They made the assumption that a calculation was being made. This seems reasonable, but if you're doing a lot of those calculations you'll memorize them. I interestingly have experimental data on this. After my undergrad I had to get an EKG done and the tech asked me to do some basic math questions to get some readings. Problem is, I could answer her questions but she got almost no signal. They were just too easy for me because I was so familiar with them. You don't need to calculate what's in the cache. So we moved to 2 digit multiplications and signal was mixed. Good correlation with being able to leverage previous calculations. So then I had her and my dad pick 3 random numbers and I would multiply those in my head. That did the trick and she said it light up like a Christmas tree (I do this visually, so it really was using more parts of the brain than she was likely used to seeing).
My point is, there's more nuance to this. Your brain isn't just a computation unit, it has various levels of storage with different speeds and capacities, it has different accelerators and processing units that can be leveraged if programmed in the right way. The problem with the OP's assessment is they've measured output speed and assumed this is enough information to calculate FLOPS, but a slower processor can win that race if it just is pulling from cache. A slower processor can win in aggregate if it has more parallelism. The problem is that they're measuring something different than what they think they're measuring, even if it is right up to a first order approximation.