Then, the question is why does Windy say the GFS resolution is 22km, not 27km?
The NOAA gives the resolution as measured on the parallel of Washington (39°N) which is about the parallel in the middle of USA. But parallels are not great circles.
0.25° measured on this parallel is not 27km, it is 22km as shown by this sketch.
This way of specifying the resolution of a global model seems to be a sort of standard now. The ECMWF center says that its model (IFS) has a resolution of 0.1° or 9km. It is obviously measured also on 39° parallel (which is not however in the middle of Europe).
Hope my sketches are clear.