A limitation of the grid method is that PSS cannot resolve two sources unless their respective peaks are separated by at least one pixel in the significance map. If two sources overlap to some degree then it may be possible for PSS to resolve them simply by increasing the grid resolution using the SAMPLE parameter. However, this method assumes that there is a saddle point on the significance surface to be resolved. If this is not the case, then the image should be binned more finely, or a photometry package such as DAOPHOT used instead.
Even if two blended sources can be resolved, the quality of PSS's results for fluxes are likely to be poor. This is simply a consequence of the model used, which assumes that the input image surface brightness can be represented by a background plus single point source at every point -- this obviously breaks down where sources overlap, where simultaneous knowledge of the contributions from two sources is required. Unfortunately, the process of fitting two sources simultaneously is considerably more difficult than twice that of the single source case, and will have to wait for a future version of PSS to be implemented.
Another aspect of the source confusion problem arises in connection with spurious sources. These occur most often when the model used is deficient in some way, ie. poor background modelling or a psf which does not match the data well, and are often manifested on the wings of bright sources (near which of course a background subtraction program is likely to have more difficulty).
PSS has a mechanism for deciding whether a source is too close to another to be real. After the significance threshold is supplied, PSS finds all the positions of all the peaks in the significance map passing the threshold. These are ranked in descending order of the distance of their nearest neighbour. Source pairs whose separations are closer than the 50 hidden MULREJ parameter (default 1) are considered to be one source. In this case the less significant (and less bright and/or psf-like) is discarded.
In certain circumstances it is possible that this mechanism appears to be merging peaks separated by small amounts. In this case the value of MULREJ should be reduced to make PSS realise they are separate. The opposite problem of multiple detection of a single source can be controlled in the same way, but by increasing the value of MULREJ. The latter problem can cause great difficulty if the source subtraction operation is used (Section 4.13).