You can trust it because it is a valid ST score for any Skylake based device at ~3.5 Ghz. I know it because I just validated it with my own system.
7700K @3.5 Ghz
7700K @3.9 Ghz
(DDR4-2133 CL15)
If you are saying this 6600K device was running at 3.9 Ghz in the Single-Core test, look at this:
7700K 3.5 Ghz
http://browser.geekbench.com/v4/cpu/compare/12915383?baseline=13335188
7700K 3.9 Ghz
http://browser.geekbench.com/v4/cpu/compare/12915383?baseline=13335375
The Whiskey Lake sample is running slightly below 3.9 Ghz which is in-line with my 3.9 Ghz test, frequency measurement from GB4 is correct there, pretty sure it is correct for the Icelake device as well.
A typical GB4 ST score for the current generation is around 1.25 X MHz at high frequencies and a bit more than 1.25 at lower frequencies.
How much of that is realized depends on the percentage of time that can be run on the max boost frequency, which is again improving over time by better power algorithms.
So I would agree that the 2015 i5-6600 doesn't run at its maximum boost all of the time.
If you now look at the image of the first GB4 compare assuming Ice Lake does run at 3,482 MHz
Whisky lake has 1.27 X MHz.
Ice Lake would have 1.50 X MHz
This is a lot, not impossible, but the reason that I find this unlikely is the correlation between 12 of the sub-test results. They are virtually the same for the two CPU's.
It would be highly coincidental that architectural improvements do benefit so many sub-tests in virtually the same way. All these 12 sub-tests would receive the same 12% IPC improvement (3.9/3.48) .....
I would expect much the performance increase to come from for instance AVX512 implemented in the libraries as shown for the other sub-tests.
The latest versions of ICC will without doubt be able for Spec2006 and Spec2017 and a number of other tests.
But we will see soon enough....