We run the tests on the 3960X because it's a "hot" chip, the idea being that if a cooler can cope with its thermal characteristics when Overclocked then it is most definitely going to be able to cope with other CPUs.
Secondly, to compare the performance of one cooler against another it's essential that all the tests are run under the same conditions and with the same hardware. Imagine if we ran GPU tests using different hardware and different test set ups and benchmarks etc. there'd be no continuity and as such no comparisons could be made which would render the whole thing pointless.
Having explained the reasons for two of your points, i'm going to let you go away and work out for yourself why the H100i results are as they are.