iraf.net

In daophot photometry of crowded fields, during the iteration to remove neighbors that blend PSF stars that cycles psf-> nstar-> substar-> psf, where the photometry input file to psf is the substar output file with unsubtracted PSF stars, nstar fits the latest iteration of the model point spread function (also denoted psf) to the selected PSF stars and revises their magnitudes. I notice that the fitted magnitudes for all the PSF stars decrease systematically by exactly the same amount every other iteration of this loop (even after convergence), although the PSF star image pixel values hardly change (after the neighbors were removed). This can be a 37% change in nominal flux with each pair of iterations (the amount depends on the choice of PSF stars). Even with just a few iterations to remove the neighbors, the error in the photometry can be significant. The choice of psf varnumber, function, datamin,max, and fitrad don't seem to matter, I am using PSF stars with low chi and sharpness with normal-looking profiles. I am running Pyraf 1.2.1 with Python 2.4.2 on a Mac OS X 10.5.8. The listed magntiude of each selected PSF star in the psf header (for example) decrements by identically the same amount on every other iteration of the above loop after convergence (where there is no change in the intermediate iterations). The amount of change seems to be (very roughly) proportional to the number of PSF stars selected, though I have not checked this thoroughly.Has anyone else encountered this bug or know of a solution? Has anyone found this NOT to occur? If so, what version/system were you using?
-Thanks