Convergence criterion. Why deltaS = mag(s(n)-s(n-1)) , not mag(s(n)/s(n-1)) ?

Why simulators like HFSS and CST use (as a default) absolute convergence criterion not relative? I guess the last method could be more accurate because it for example can make difference between s11(n)=1, s11(n-1)=0.9 and s11(n)=0.2, s11(n-1)=0.1. In the first case relative error is 11% in the second - 200%, but absolute error is constant = 0.1. Why so?