Images from the Hubble Space Telescope suffer from an overcorrected spherical aberration that is due to a conic-constant error in the primary mirror. Within the program known as the corrective optics space telescope axial replacement (COSTAR) simulators have been built to provide the point-spread function (PSF) of the telescope alone and of the telescope with the faint-object camera F/96. It was found that the experimental PSF's were identical to those in orbit, which was not the case when the PSF's were calculated with commonly used optical software. We explain this discrepancy and propose a modeling method that is based on the determination of the wave-front error at the exit-pupil level that gives results that are consistent with observations.