what XP edge is the probe touching the core? One thing you have to realize is that a side-mount thermistor isn't exactly the greatest cookie in the world....since it also measures pcb, core, air around the cpu, grease plays a role, etc.
That said, socket-thermistors are junk for anything other than pure-fire alarms, but to function properly as a fire alarm they must be calibrated for each cpu/heatsink/mb combination.
Btw, here's a good watercooling roundup, with very good testing:
Pro-Cooling water cooling roundup part 1.
They use a method sanctioned by amd (24228.pdf), with a behind-center core socket-thermocouple. AMD also provides a formula in the pdf to negate the thermal resistance of the ceramic and calculate "die" temp from this backside thermistor/thermocouple:
Tdie = (1.209*(Tceramic-Tambient)-1.3778) + Tambient
Granted, this is a ceramic PCB T-bird, and not an organic PCB XP chip(amd does not recommend or even mention using socket-thermistors for XP chips

), but considering that the placement of the 8kha+ socket-thermistor is even worse than the abit kt7, and the pcb makes for even poorer thermal conductivity, its a bunk comparison, but this info can still be useful. if you take the first waterblock, you end up calculating the "die" temp to be around
49C. strikingly similiar results abound the rest of the review too. The general conclusion: The Socket-thermistor, despite the lack of airflow, still reads lower than core temp. Its generally much closer than with an air-cooling method, though.
But your temps are great, regardless. Its just that these chips really run a lot hotter than a lot of people realize. Just look at this review from Void Your Warranty. The Thermaltake Volcano 7 was capable of cooling a 90W processor to a theoretical "Die" temp of
74C.
Void Your Warranty Volcano 7 Review.. Notice hwo they get side-thermistor temps under MBM temps.
Mike