I unpacked the chunks and there doesn't look to be much more (or it's subtle).
The IDAT chunk (which makes up most of the file) is DEFLATE-d but uses full color (rather than palettized color, which would have saved lots of space, and might be how you'd get 40% compression). It also doesn't use any row filtering.
I checked incase something was hidden in the DEFLATE encoding itself (e.g. using sub-optimal block encoding selection) but this round-trips exactly, so I don't think it can be.
Edit: You're right though. Using imagemagick to clone it like so:
results in an image that's at least 40% smaller. It might just be better compression, but there might be another layer of steganography going on.