Criticism of statistical energy analysis for the calculation of sound insulation: Part 2-double partitions