There are 21 stars in the overlap region between frames, on average.
The error in frame offset would be 0.2 pixel if one simply differenced the source positions in the overlap region. This would be 0.06 pixel rms.
A simple random walk with this as the step size would result in the position of the last frame have a sigma of 0.06 pixels relative to its true position, where 259 is the number of frame differences in a scan. Note that the previous note from T. Chester failed to account for the correlation in the frame offsets, caused by:
Another way to see this is to realize that of the 5 position deltas between the 6 occurrences of each star, only the first one is a brand new measurement (x2-x1), with all other position deltas simply replacing the second position with successive positions.