Wind-shake, astigmatism, etc., can cause differences in image shape (e.g. orthogonal elongation) which cannot be transformed to a common shape using convolution. In these cases, our image subtraction software will detection residual subtraction errors and flag these as candidate supernovae. The task is to develop diagnostics, from the input images or the subtraction image, so that these spurious candidates can be flagged. Since subtraction errors are highly correlated, it would be best to define search exclusion regions. These regions will have to be cataloged so that we know exactly where we have searched.