no thermal compound difference make a 3 or 4 deg C difference. There are too many variable involved to trust data like that. To truly be scientific about it, a measured layer (to manufacturer's specifications) would have to be applied (not something you can do by hand). The pressure and planarity of the heatsink would have to be regulated and measured (to eliminate differences that occur with minor mounting variations due to tolerances). Then a true caloric measurement of heat put out by the test surface would have to be compared by the true heat at both sides of the interface. This would require instrumentation that can adequately measure it.
In the real world, when you use a software tool that reads out the hardware that is on a motherboard you are not getting precise or even necessarily accurate information. Don't be misled by data, like the NVnews link pointed out. You would be amazed if you had ten identical computers, and saw what the variation was in temperature readouts on each one of them while running the same benchmark. There is around a ten percent difference between them all.