I use a calibrated, old photographic mercury thermometer with increments of 0.1C on the scale to check my sensor temps. I know it's not a digital device but the scale is large enough with small enough increments to read with great accuracy. The only problem is with positioning to get a reading so I am aware that the data I obtain from it may not be precise. It is however, extremely sensitive and will react even just breathing on it or placing a warm or cold object near it. I calibrate ambient air temp sensor, internal case temps and coolant sensor in reservoir while computer is at idle according to this thermometer. I find that as temperatures rise with increased load I get errors on my sensors up to 1.6C below what is reported on my thermometer. It is always below, never above. I don't usually worry about it because it is all well within the thermal limits of the hardware and 1C or 2C doesn't amount to anything in terms of my 'performance needs', however ... I have not monitored any possible variations over time with my sensors or seen inexplicable jumps as mandrix has seen, but then I'm not really looking. As I haven't played around with the data logs much I might try it on the sensor temps and see what turns up.

I don't know much about these particular types of sensor but I imagine they probably wouldn't be linear in their scale? What is their effective temperature range and how do they behave at the extreme ends of their range? Are the sensors consistent or do the scales vary between sensors? It would be an interesting project for someone in the watercooling community to set up a controlled experiment into the accuracy of these sensors across a temperature range. My penny's worth, hope it assists just a little?

I find positioning of ambient sensors is critical. Even if placed right in front of intake fans the slightest backwash or unpredictable air flow can cause readings to jump around. I use multiple intake sensors in different locations and combine them using a Lowest temp Virtual sensor that gives me a quite steady ambient.
I chart my ambient, water and delta temps and any spiking tells me that something isn't positioned correctly or is being affected somehow. I mostly see nice smooth lines and curves.
My water sensors don't jump around at all and they are exactly the same things under the plastic cover.

I have noticed that one of my sensors in particular was jumping around by up to .4C for no reason but I swapped it out and the problem hasn't recurred that I have noticed.

Just an example of my chart at idle

Yeah just in the Data log tab add a few data sources, set up and save a chart or two in the chart manager then add add a Log Data Analysis object in an overview page and link it to the chart you created.