Calibration is best done against a primary reference, failing that a more accurate instrument than you are trying to calibrate can be used. At "normal" temperatures an Ice bath and a hypsometer makes reasonably accurate 0 and 100 deg points. Beyond that a "triple point cell" is used - making your own IS possible, the amateur scientist column in Scientific American had an article about it two or three years ago.

Above that range, you usually use the freezing points of pure metals as references.

Steve