A point that I have not read yet, is the other secondary factors like airflow mass movement that is supposed to equalize the temperature in a closed environemnt.
I noticed that in every house that I have lived (over 5), there is always that hot spot or room that everyone gravitates towards. ;-)
Hahaaa!! But pugsley does make a very good point. What is comfortable? And this will have much to do with one's metabolism and circulation. Under normal conditions of physical activity and diet I can estimate the air temp to about half a degree F. I'm presuming here that the walls are also about the air temp, so the radiative contribution is consistent.
But a touch of illness or a sharp reduction in fuel food and I'm feeling cold in ambients of, say, 73 F. And my estimates of temperature are way off, unless I knowingly compensate.
Long ago I did weather observing in the Arctic using liquid in glass tube thermometers mounted outdoors. We were supposed to read the liquid to within 0.1 degree (try doing that at 40 below and your eyes streaming from the wind blasting). Each thermometer had a serial number and a conversion chart for correction.
If I recall correctly the reading error could be as large as 0.5 degrees (can't remember if F or C) and the chart had a curve that indicated reading vs true corrected temperature. This was long before electronics was accurate to 0.1 degrees F.
I bought a "scientific" thermometer years ago which has a long stainless probe that protrudes, and has a resolution of 0.1 degree F. It seems to have kept its calibration over the years, although I can't test it with boiling water as the highest reading before it overranges is 140. A ice and water slurry would be at least one check though.
When my late mother once stayed with me when I rented a guest house in Malibu, she was complaining of being cold until I used some electric heaters (I think for the first time) and brought the temperature up to about 85F according to the scientific thermometer. Insanely hot for me, and I pointed out to her how warm it measured. She said Oh that's ridiculous --- it's not that hot in here! And besides, that's a MEAT thermometer! :)
I've found the National Semi temp sensors like the LM35 to be at least in good agreement with the scientific thermometer.
Another approach that I saw in the Review of Scientific Instruments years ago was to separately measure a resistor's open-circuit voltage noise, then the virtually-shorted current noise, and back out the temperature (needless to say the bandwidths have to be accurately determined, and the noise of the preamps similarly well-characterized). But the cool thing (indeed this was presented as a technique suitable for cryogenic apps) is that the actuall resistance value drops out in the analysis.