Well, the rest of my hose barbs came yesterday, so I fabricated a modified hold-down for the other modified graph-o-matic and took the machine down to install it. Now both outlets of the WW have an identical chipset block on them, and my CPU temp is back to a pretty consistent 75F. I didn't remount the WW, but I did put another turn or two on the nuts to compress the springs a little more as Cathar suggested.
I still don't know how much I can trust the socket thermister. Here is the low-down on temps and load right now:
CPU temp: 76F (socket mounted sensor read with SpeedFan)
rad intake air temp: 75F (indoor/outdoor thermometer's indoor sensor)
rad exhaust air temp: 80F (indoor/outdoor thermometer's outdoor probe)
rad exhaust water temp: 82F (capillary-style photographics thermometer submerged in res)
CPU: Palomino XP1800+ (1533mhz; everything is running at stock speeds)
CPU usage: I'd estimate 20-30% average over the last half hour or so (I'm installing software, browsing web, and playing music)
CPU TIM: Arctic Silver 3 paste
There is also a GF4ti4200 and my Abit NF7-S's nForce2 chipset in the water loop, both using previously described
modified Innovatek graph-o-matic blocks.
Which temps are most suspect? Probably the CPU termister? Anyone know where can I find info on how to read the internal die sensor on this CPU with the NF7-S? I'm having a rough time finding info on this.