I am not a complete expert on this either, but I have written quite a few converters in my life.
@kh99 has it exactly right about the png deflate compression. In most cases these are back-end optimization settings and there is really no need to have your users set these values.
To sum up, on thumbnails... If I use image-magic/mogrify I use 85% on the thumbnail. GD2 I use 75% since there doesn't seem to be any great output clarity difference at 85%. if you detect retina devices and you have the ability to store separate thumbnails, then the numbers above my not apply, but I use the same numbers.
Portable devices we display a special resolution picture and no thumbnails, I played around for a long time on multiple devices and found that between 75% and 85% seemed optimum. Retina of course we send double the resolution. We use this lower sized picture as our thumbnails for desktop and let the user set the size of thumbnails. We have never found a need to alter the loss-less settings after our initial determination.
PS: Our pictures are typically technical so clarity is everything. Make sure you server does not try to re-deflate the images.
You might also want to look at 'jhead' and 'jpegtran' if your trying to add various user tools for your images.
Have fun!