E6600 on watercooling/Accuracy of temps

Sort of a longish two-topic post: First, an examination of the temps I'm getting from my water-cooled E6600 at 2.997GHz. Which leads to the second topic: how to calibrate all the temp monitoring software to get really accurate temp readings.

First, for your comments and opinions, my temps, as read directly from the applications referenced below, uncorrected. Temps are given in deg. C and are in boldface.

Variables
Ambient intake temp: 29
Chipset: Intel 965P
CPU: E6600
CPU Cooler: Swiftech H2O-120 kit
Frequency: 2.997GHz, 333 clock X 9
Load: TAT 100%, but note I will also give idle temps first for a baseline
Motherboard: Gigabyte GA-965P-DS3, version 3.3 (F10 BIOS)
Vcore: left at normal, no increase in BIOS. Speedfan reading given below.
Case (if it matters) : Lian-Li PC-A10A

NOTE: Contrary to parameters in Temp Guide sticky, I ran the fans at reduced speed during these tests because I am shooting for a very quiet/silent computer.

Fan settings:
Front intake fan and card bay fan (card bay fan is reversed to work as an intake now) on "Medium" setting of case's built-in controller. Approx 7.5V
Blowhole fan (case fan, running as exhaust) using 7V resistor cable
Radiator fan: 5v

Temps, after stabilizing at IDLE for 10 min
TAT: C0 32, C1 29
SpeedFan: C0 30, C1 29, T1 (presumably the MCH) 45, T2 (Tcase?) 20, Vcore 1.31volts
Note that the T2 temp, which is probably the Tcase temp (it increases very rapidly when a load is placed on the processor), is reading as below the ambient temp. of 29 deg. C.

Temps, after 10 minutes at TAT 100% load
TAT: C0 53, C1 50
SpeedFan: C0 54, C1 52, MCH (T1) 47, T2 (Tcase?) 42, Vcore 1.28volts

After stopping the TAT 100% load, the C0, C1 and (presumed) Tcase temps drop by 10 deg. C in less than 2 seconds, and continue a fairly steady decline at a slower rate for a minute or so. They ultimately reach the idle temps given above, but the last two or three degrees drop takes two or three minutes total. I'm guessing that the rapid initial drop reflects the rapid heat removal provided by the water cooling system, and the slower drop at the end of cool-down reflects the "stabilizing" effect of the small amount of heat already built up in the coolant, which cools down more slowly.

Now, the second part of this post: Obviously my Tcase temp is way off, as it shows a temp that is below ambient. Oh, yes, I also ran the Gigabyte Eazy Tune 5 software only to use as a temp and voltage monitor (I know better than to use it to O/C, that is done in the BIOS), and ET5 showed a "CPU" temp that matched the T2 temp from SpeedFan, so this tends to confirm that T2 is Tcase.

If my ambient is 29 deg. C, how do I know exactly how much to add to the temps that the software is showing for Tcase? Should I take the middle ground of the range given in the Temp Guide, ~1 to 15 deg above ambient, so add about 7 or 8 deg C to the Tcase reading? But that would still give a Tcase nearly at, or slightly below ambient.

And how accurate is TAT's core temp reporting? Are the idle temps and load temps it gives accurate? Does watercooling work so efficiently that the cores are actually at 32 deg or so when the ambient is at 29 deg? That would require not only a very efficient cooling system, but very efficient heat flow from the cores to the heat spreader.

It seems to me that without actually taking temp readings directly from the heat spreader and cores, it is almost impossible to calibrate for the errors in the whole, multi-stage (da diode is connected to the firmware, da firmware is connected to da BIOS, da BIOS is connected to da software, dem bones, dem bones, dem dry bones), system that we all use to monitor our processor temps.