I have been doing a bit of work. I have replaced some thermistors, recalibrated, remeasured the fluxblock after some lapping, fine tuned the placement of thermistors within the fluxblock and tweaked my realtime calculations to take into account the heat shadowing effects of the sensors and their holes.
I am now fairly confident that I am able to get reasonably consistant readings from the setup.
Since I generate a waterblock C/W number by extrapolation, I am able to get a number independent of the value of the TIM by assuming that the TIM joints at either end of the fluxblock are the same since I measure and calculate the joint between die and FB and remove it from the WB number.
This relies on me getting a good mount. Since I have made many mounts of this waterblock, I am now pretty familier with its performance so I am able to see if I have made a good mount by checking that my calculated C/W is correct. I tighten the mounting screws under power and am able to see the measured WB and TIM C/W in real time while I am doing this. When the WB C/W is correct and the TIM is at a minimum (within the mounting pressure limits, a spring compression distance) I know that I am either good on both TIM joints or equally bad on both. I choose to believe that it is unlikely that one would get both interfaces equally bad at the same time. I can see a difference - if the WB C/W is high, then the WB-FB interface is bad, if it is low then the Die-FB interface is at fault. This relies on a bad joint being caused by uneven pressure tilting one of the interfaces. Anyway, details...
(This reads incoherently, I will fix it tomorrow)
This shows my performance over six mounts. (It's actually good over a damn sight more than just six) The Shin Etsu 765 grease (and 751 actually) is extremely viscous and seems to go hard when you move it or put it under pressure. A non-Newtonian dilatant behaviour. I had terrible difficulty in getting any decent mounts at all with it. This shows in the chart. Apart from this one thermal compound, I am pretty happy with the repeatability I am able to get.
The following numbers represent what I have been able to get from several different thermal greases. I would caution people not to draw too many conclusions from this. The numbers are Short term values and do not show effects of TIM settling over time and power cycles. They also represent only the behavior under a given pressure (~20kg over 144mm^2), NOT a given thickness. Meaning essentially that the more viscous the grease, the higher the thermal resistance. Depending on the application, the value will probably change. (Think IHS, where there might be a measurable gap between die and spreader)
A simple chart for what is tens and tens of mounts, megabytes of logfiles and repeated remounts.
Edit: Manners. Big thanks Colin at Arctic Silver for providing the Arctic Silver greases and Bill for the Shin Etsu and a pump, and the waterblock.
Excel sheet attached for those interested.