I would agree that the source image Gib used for that one does seem better to begin with. Another factor to bear in mind is that the palettes in my images have been converted to match the OCS 4096 colour palette as closely as possible, so hopefully that should allow more faithful colour reproduction. I'm not sure if he took this extra step. Did Gibs mention which dithering algorithm he used?
To be honest I do not believe that my chosen dithering or quantization algorithms are the issue here (Sierra Lite and Wu v2 for those interested in such things). And whilst some images already produced will look better, others won't. Reducing colours from true colour to 16 colour palettes doesn't always produce pleasing results and often people result to tweaking by hand to get the best results. Not really an option for such a large image set. But hey, we can try different algos if there's interest.
