Figure 6.

Type II has multiplicity one and is a descendant of type I. Thus all pairs of mutations that do not include the outmost mutation of type II weigh the step removing the outmost mutation of type two with zero. The table on the right shows that $Q_{\theta }^{\mathrm{\Lambda }­{\mathrm{HUW}}^{2}A}$ is closer to the optimal distribution than $Q_{\theta }^{\mathrm{\Lambda }­{\mathrm{HUW}}^{2}B}$