If you can be assured of linearity the water melting point, boiling point calibration is pretty damned accurate. The tricks are:
The water/ice must be pure, as in distilled.
Use slush (crushed ice) with the water level slightly below the ice. When determining melting point remember that water is densest at 4C so the ice/water mixture must be continuously well stirred or the water at the bottom will be 4C and the rest of it somewhere between 0C and 4C (magnetic stirrer works best). A steady well-stirred temperature is 0.0C.
For the boiling point, use glass beads (small stream pebbles would work) to prevent bumping of the ultrapure water. Use an erlenmeyer flask (or something else with a narrow neck) with the thermometer clamped to a stand and the thermometer mounted ABOVE the water in the steam space, arranged so that it doesn't get splashed with water. Record the steady steam temperature at a rolling boil. Record the barometric pressure.
You can find charts of boiling point vs atmospheric pressure (
http://www.hi-tm.com/Documents/Calib-boil.html) on the Web or use the steam tables (look up saturation temperature for corresponding atmospheric pressure.
http://www.efunda.com/materials/water/steamtable_sat.cfm ).
