Attached is the pic Roscal provided about the IHS's.
The IHS thickness difference is pretty substantial between AMD and Intel. Looks to be 1mm. Some water block bases are only 1mm thick and for a reason. The thicker you make the IHS the less the thin bases design blocks will perform.
Now if you remove the IHS you get rid of that extra thickness but you also remove that thickness variable from the test which IMO is not good. You basically remove any idea how the block will perform with an IHS which 99.5% of the world will simply not remove. Not to mention they probably will become non-removable in the future anyway.
IMO if we really want the best idea how a block will perform on a actual CPU then we would have to come up with a way to test with the IHS and take into account the various IHS thickness's.
As for using IHS temp or die temp I don't see that it matters. I would prefer the temp probe NOT touch the water block in any way however. Cooling the probe is probably not a good thing. Closer to the die the better IMO.
That is my confused take on it. The more I listen the less I learn lately though.
