That can't be right, all reports I have read about comparing xz with gzip tells me that xz will have better compression but at a slower speed than gzip.

Have you run a test of gzip at the same compression level with the same data?

I don't doubt that xz will be smaller, I doubt it will be quicker.