I've run through a little accuracy test of the sun position code I'm using right now, as provided by Mowcius. To check the accuracy I compared results calculated by Mowcius' code and the NREL online solar position calculator solpos (uncertainty of +/- 0.01 degrees).
Results were calculated every 10 minutes for the 11th of November using location data - Longitude = 172, Latitude = -39.
Here is the plot showing the error during the day:
Turns out Mowcius' code is not extremely accurate. But the errors are somewhat consistent and the results for each calculation could be improved, for this day at least, by adding constants (Elevation -2deg and Azimuth +4deg say). Hopefully I can track down a more accurate position calculator, but right now at least I know how much error to expect from the calculation.