A very good discussion of this has been provided by Otto Zork. I have added some of his points and ways of expressing them to this discussion.
Steve Havas got a series of images of Nancy's coordinates on Sept. 21, 2002. Nancy, Charles, Steve, and Jan have been trying to guess where her planet might be on these images. In fact, all of the "stars" they have picked out turn out to be pixel defects on the images. This is not surprising since they were constrained to trying to find things on Steve's images that were not on preexisting images of the same area of the sky.
One of the Nancy selected "stars" in Jan's summed stack (this is the one she has selected as the real thing) is a single 4 pixel defect on shavas-5-R.fts:
Another "star" is a single 2 pixel defect on shavas-15-R.fts (there are two more defects here not yet pointed by Nancy and her followers ... perhaps they will do so later):
Note that the characteristics of the pixel defects are very different from those of the astronomical objects in the field. Pixel defects are compact, rectangular groups of a small number of pixels. The stars are larger, they "fade" out towards their edges, and they are somewhat elongated because the telescope was not tracking perfectly during the exposure.
Jan has suggested that IMO has used an incorrect method of combining "stacking" images in order to hide Nancy's star. In fact, the median combine used by IMO is used by CCD imagers because it allows one to exclude defects from the stack while retaining the real objects. This is illustrated below:
The full set of 20 R images, stacked by the "sum" method (Jan's prefered method):
The set of R images, less the shavas-5-R.fts and shavas-15-R.fts images containing the defects shown above, stacked by the sum method:
The full set of 20 R images, stacked by the "median" method (preferred by IMO):
Notice that the median combine retains all the stars seen on the sum, but does not have the potentially embarrassing (if you announce your results to the world) defects.