Lumped convolution integral models revisited: on the meaningfulness of inter catchment comparisons