Even better when you switch on the --lossy option in gzip.
Admittedly, it's somewhat unclear why the format didn't take off. At its core, it uses the Lessis-Moore algorithm to perform a global bit-wise sort of its inputs, then performs a run-length encoding (plus some other magic) on the results to achieve up to 100% compression.
[1] http://web.archive.org/web/20050402203136/http://lzip.source...
I'm not convinced that actually qualifies as "lossy".
gifsicle --lossy actually allows slightly lossy LZW dictionary matches to improve the compression itself.
If gzip tried to do lossy compression, not knowing what the data is, it would have to randomly corrupt data.