Application of end‐member modelling to grain‐size data: Constraints and limitations