Agreed, there are two kinds of inaccuracy at play here, one is the linearity and the other is absolute reading. The linearity shouldn't be a problem, it is a measure of how much DIFFERENCE there is between the voltage returned for 1 degree change when near its cold limit and 1 degree near it's hot limit, as you are measuring over a small range it will be insignificant. The absolute measurement is the one that might be different from one sensor to the next but it's easy to trim out the error as it will be a constant.
Brian.