I agree. Given the rounding (43.0 and 42.9 could theoretically just be a 0.01 difference) & measuring bias there might be no declining effect here at all.