“…We can also assume that C G (s) is not a torus. As σ 1 interchanges the long root subgroups of G with its short root subgroups, the tables in [55] only leave the possibilities (1, 1), (4, k), (14, k) or (15, k) for the G-class types of s. As σ interchanges the maximal tori of type (20,12) and (20,17), it also swaps the classes of type (14, 2) and (14, 3) and the classes of type (15, 2) and (15,4). Let us now prove the remaining statements.…”