Let's pretend for a second that the results are accurate to 0.1 degrees. The improvement given by the 947U over the 900U is equivalent to changing the 900's mounting system. That's not exactly a big jump. In fact, all three heatsinks are practically identical, and 1 degree variation could be observed simply by re-installing the same heatsink over and over.
Here's a self-contradicting quote from the article:
The rise-above-ambient load temperatures show a statistically insignificant 1ºC variation, but the 947U does in fact come out on top[.]
Even the authors admit that there's no significant difference between the three heatsinks, although for some reason they still claim that the new heatsink is better. I guess if you spend that much time testing a heatsink, you want something to show for it. You can't just tell your editor, "Yeah, they're all the same heatsink," you have to have a winner!
If anything, this article shows that cooling has not
improved with the new heatsink.