Often the grades will follow a gaussian (normal) distribution. In that case, the wider, less used range, is important in capturing and quantifying the outliers. It would be an effective grading system if it saturated at the high or low end because it would no longer have the power to differentiate at the saturated end.