The simulation results were checked in many different ways to verify their accuracy. They correctly reproduce the theoretical results if one simply uses the differences in the source positions in the overlap region (that was how the correlation was discovered!) The final reconstruction produced a reduced chi-squared extremely close to 1, implying that it did the best job that could be expected in modelling the data.
All offset reconstruction methods gave similar results, which were about 15% better than the simplest method of differencing the stars directly from adjacent frames.
The position errors, using the PPM stars to determine only a single offset for each scan, are of course at a minimum at the center of each scan and highest at the end of each scan, varying quadratically from the center. (This is because the ends of the scan are on average farther from the PPM stars.)
The table gives the positional accuracies, including an extraction error of 0.16 , resulting from applying a single offset to the entire scan derived from all the PPM stars in that scan.
Note that these results are comparable to the accuracy of the PPM catalog in the North, which is claimed as 0.34 (the value put into the simulation)! The accuracy of the PPM catalog in the South is claimed to be 0.14 , so the results should be better in the South.