Relative alignment was performed by finding the peak of the cross-correlation between images and a pointing master selected from the epoch with the best-constrained pointing model for that field. Each observation was initially mapped individually, then all observations of a given field were cross-correlated with a selected master image of that field. The cross-correlation peak was fit with a gaussian and the difference between the gaussian peak and the image center was used as the pixel offset. The offsets were recorded and written to the timestreams. Finally, all observations of a field were merged into a single timestream with pointing offsets applied to create the field mosaic.