I think the range determined by fit.variogram is not the "practical" range as mentioned in Applied Spatial Data Analysis with R. Do I have this right? To get the range used for interpolation, I adjust the range (eg. 3^0.5 * range)?