Following intel's thermal guide to measure the temperature on CPU surface, we usually drill a small hole right at the center of the heat-sink, and then insert a 36AWG thermocouple through this hole. There is about 2~3mm of the thermocouple junction porption resides in the heat-sink's base plate, where,you know is a high temperature grident region, and the thermocouple surface is surrounded by a boundary condition of high temperature grident. Dear Sir, do you think this will lead to serious measurement errors, except conduction ? Many Thanks !!
It is not easy to predict the error that we could possibly get, but considering the accuracy of the board (you did not mentioned what board you have, but you can go to http://ni.com/advisor/accuracy/ and take a look at how accurate it is) we must have pretty good readings.
Now, regarding the physical phenomena that you talk about, there's two things that we advise you to do: 1. make the thermocouple junction as small as possible, so it gives more accurate readings 2. try to isolate the part of the system in which you want to measure the temperature. This can be done using glass or any other good isolator.