The code would previously only consider the last chain mapping in
consideration. As a result we find better scores in tests.