Statistical treatment of grain-size curves and empirical distributions: densities as compositions?