Sorry for dragging up an old thread but it's really good info here.
I use a blow-off tube on every brew, and feed the probe into the starsan bucket, through the tubing, down the airlock and into the wort. After a few brews I got paranoid that I was poisoning myself from the plastic probe (stupid, I know) so I got some food/medical grade shrink wrap and put about 15" on the probe.
I converted a commercial Pepsi fridge with a Love controller. I wired the fan for continuous operation and use the controller (along with a SSR) to cycle the compressor. I tried taping it to the bucket w/ insulation but wasn't convinced I was getting an accurate number so I too various readings around the fridge and inside actively fermenting wort. Vigorously fermenting wort was a full 8 degrees away from the probe reading (I calibrated before experimentation) and fell off as the yeast settled down.
I think a lot has to do with my choice of chambers, and that the commercial unit was a poor choice, but hey the price was right. Because it's a beverage cooler it's cooling power is way, way oversized for my application, and was leading to short-cycling. I compensated by increasing the differential to 3 degrees but even this was giving me 3-4 cycles per hour and the outside temp was not linear with the actual wort temp.
Long story short, with the probe in the wort the compressor cycles 1-2 times/hr for about 12-15 minutes and the wort temp stays within 2 degrees of target.