The Ice Cloud and land Elevation Satellite (ICESat) has been acquiring unprecedented observations of our planet's varied surfaces with impressive precision. However, systematic errors remain in the data including orbital variation and long-term bias trend pointing errors. These errors are particularly troublesome because they cannot be separated from true surface elevation change and confound instrument range bias determination. We present a method to correct these systematic pointing errors to the sub-arcsecond level. We demonstrate the impact of these errors and their corrections on Greenland and Antarctica ice sheet surface elevation change observations and sea surface anomaly observations, as well as instrument range bias determination. While further improvements in the ICESat data will certainly be made, the analysis presented here demonstrates a major error source can be significantly reduced, resulting in improved surface elevation accuracies for ice sheet elevation change detection.