[thearn4]

I think it's because the deconvolution is a bit tricky (with a few different "knobs" to turn somewhat arbitrarily in the process), and can in some ways be expected to add in more artifacts than it is subtracting.