> All this is to say: yes the data are noisy, but we'd need to erase/explain away the numbers by at least an order of magnitude to erase the spike we are seeing, and that seems like a tall order.

That's fairly easy: https://en.wikipedia.org/wiki/Frequency_illusion