Is it possible to read the strength of the hall sensors? (now I see only 0/1)
If so these could be used to interpolate between the values. The 16 wind directions are steps of 360/16 = 22.5, if interpolation is possible you may could get to 5 degrees accuracy