First off it's not clear if you or warmmilk meant cpu - ambient or pure cpu temps regardless of the ambient. I've seen both out there in a variety of reviews.
CPU temps will raise as the water temps raise. The delta should be maintained within a reasonably small error. I'm sorry if you took my comments personally, but feel free to correct me with reasons if you disagree. I would still argue that measuring vs coolant temperature is more accurate than measuring vs air. It removes unnecessary errors and variables that need not be there. I'm not trying to say that there are no effects, simply that those are smaller than those would be vs ambient. Also bear in mind that when you vary flow you're also varying your radiator performance too. If you haven't calculated that in when you measure vs ambient you quickly start measuring the system performance and not the block performance. Yes maybe you could get the same accuracy by perfectly controlling and measuring ambients and keeping your radiators clean and subtracting radiator performance vs flow. But even then, you would be more accurate if you measured vs coolant. If you can be more accurate, why would you choose to measure with less accuracy?
Personally I think cpu-coolant is the most accurate way to measure the block performance. However I also know from first hand experience that it isn't perfect. There are secondary effects such as cooling through the socket and common mode shifts with temperature due to presumabley poor cpu temperature linearization. Airflow blocked in a radiator will cause a shift in coolant temperature which will stack on top of any ambient variation. This can be minimized by oversizing your radiator and by minimizing ambient shifts. This is all much harder and needs to be much more tightly controlled when measuring vs ambient. That is my point, and I think your point is that if you control those well that you can still be accurate and this is true if fix the flow rate. But why not be more accurate?
As for issues of digging at others - that was not my intent. Ok maybe ignored was too strong word, and it seems like there might be some butt hurt going on, but if a reviewer is choosing to be less accurate when they could easily be more accurate then I have to question what else they are taking short cuts on. That's the point I was trying to make by saying that because there are plenty of reviews showing data at 0.1C resolution when their ambients aren't even measured but are "controlled' to an unspecified resolution and flows are not fixed. So if I've offended you then I apologize, but nothing you've said has convinced me or demonstrated that measuring vs ambients in the same setup is ever as accurate as measuring vs coolant temps.