It’s just a rough guess based on the redundancy in that image (limited range of colors, no hard edges). Lossless compression ratios can vary wildly depending on image contents.