Due to the errors in astrometry caused by the bright star effect we need to reprocess the STS observations from 2010. A first pass using the STS_DATASET recipe developed by Gene did noticeably improve the results, but there are still significant errors.

After some iterations we have come up with a set of recipe values that seems to generate more acceptable results for the troublesome chips. 

Examining plots of dRA versus instrumental magnitude we find that the bright star effect occurs for stars brighter than -12. On the other hand at around -11 the catter starts to increase significantly. If we restrict the astrometric fit to use stars between -12 and -11 we get better astronomical fit. 

The attached image is a montage of plots of the difference between the RA of the detected stars and the reference catalog versus the instrumental magnitude for exposure o5426g0113o. We see now that for stars with instrumental magnitude greater than -12 the scatter in dRA is centered on 0. 

More testing is needed to verify the results.