For all those that think having the temp gauge sensor in either cooling line is the best place, you should know that the majority of the trans fluid does not get sent to the cooler each time it cycles through the trans.
I agree, but the pan will also not tell you the actual max temp that the fluid reaches since much of the fluid in the pan (especially in a deep sump pan with larger volume) does not cycle immediatly or withing the same timeframe you referenced either.
However, since the fluid that does go to the cooler just cycled through the pump and convertor, won't this provide the reading for the actual max temp the fluid has reached (the sender does not need to sample all of the fluid, just some of what just came out of the convertor).
Based on this theory, I would think if you want the more steady overall "average" of the entire trans, measure at the pan. If you want the max temp of the fluid, measure at the cooler feed.
My thought has been to watch for a max condition, not an average one. In a similar way of thinking (but not quite as extreme perhaps), consider gauging RPM where you do not want to exceed "X" RPM. The average RPM over the last 10-20 seconds is really irrelivant.
IMO, cooler return reading is really only good as a nice to know or intersting observation of cooler effeciency for the most part, and would certainly not be my choice if only one location is sampled.
It would be interesting to see all three locations data-logged on a graph to see how they trend against each other. I'll have to see if RanSil can rig something up like this since he is a data geek that way (he has multiple wireless thermocouples in his electric smoker and logs temps for BBQ ribs and pork butts). ;D