I moved a raster by 0.5m. The result reports max/min raster values that differ from the original. Is this just some calculation inaccuracy? I did some random sampling and the values are as expected. Also when using the Shift tool with a GDB as target, the raster values end up as #QNAN - it only works when writing an image file (like TIF).
ArcMap version is 10.5.1.
performs no resampling, so I would be surprised that the values changed unless the reported range of the outputs has been clipped somehow.
tif would be the preferred output since the coordinate system (and hence extent) would have to match if you are saving to a different location and type.