A thermowell is the most accurate liquid temp reading you could get, because the yeast during primary fermentation generates heat, so there will be a variable offset between the actual wort temp and a separate liquid container's temp depending on yeast activity, as well as the air. A separate liquid container is better than air, though. You wort could be at 72 and the air or other water container could be at 65. For a kegerator, a separate liquid container is ideal, because you would not want to bring a warm keg in and have the freezer freeze the other cold kegs trying to pull down the temp of the warm keg.
With all due respect, I generally disagree with most of the above.
1. IMO, a controller probe placed in a thermowell in the fermenter will result in larger temperature swings than having the probe in the air and monitoring the fermenter with a separate thermometer. The freezer will overshoot and undershoot a lot.
2. A controller probe in a container of water will also increase the response time resulting in overshoot/undershoot. All this does is increase the differential which you can adjust on the controller itself, so no need for a water bottle.
3. Introducing a warm keg of beer to the freezer will not cause the freezer to run constantly trying to cool that single keg. The heat transfer rate from the keg is not that fast. This just does not happen. I did have some keg freezing problems before I installed a fan. This happened because the poured beer was often too warm, so I kept turning the temp down eventually causing ice to form which blocked the dip tube. That seemed strange, but I finally figured out that the beer warmed as it passed through the lines and taps even though the kegs were colder than they needed to be.
4. IMO, trying to monitor the fermenter temperature with the controller probe is not the way to go. It's much better to have a separate thermometer to monitor the fermenter temp.
5. Attaching a thermometer probe to the side of a fermenter and isolating it with some insulation works really well. The temperature holds very steady within one degree or so this way and I believe that it provides a reasonable approximation of the actual wort temperature inside the fermenter. Using a thermowell may be a slightly more accurate method, but not worth the hassle IMO.
This is JMO, nothing more, but it's based on several years of experience and experimentation if that counts for anything.