> I still wonder how they can know that reliably.

I think the general conclusion is that they don't know it reliably; notice the other comments in this thread pointing out that the numbers jump up and down more or less every time they're measured.