Question says AVERAGE CASE time complexity only to consider .
Selection sort uses about (N2 / 2 ) comparisons and N exchanges on average.
Insertion sort uses about ( N2/4) comparisons and (N2 / 8) exchanges on average.
Bubble sort uses about (N2 / 2) comparisons and ( N2 / 2) exchanges on average.
So if exchanges cost nothing and only comparisons count, then insertion sort way better than the others.
If exchanges count but comparisons cost nothing, then selection sort beats the others asymptotically.
Hence they need to use Insertion sort for S1 and Selection sort for S2.