I haven’t used CM Switch in a while but I remembered that CM has exceptionally good documentation!
The manual (p. 96) says that in VA mode “[t]he addressing voltage is scaled such that 10 volts always equals the last step regardless of the selected length.” I assume that in length-2 mode, 0v to 5v present at ADDR would switch to 1 on each clock pulse, and 5v to 10v would switch to 2 on each clock pulse. That’s consistent with your observation that 0-5v doesn’t switch to 2, and if Permutation is doing 0-8V, you’d expect to see output 1 more than output 2 (about 2/3rds more often if the voltages are uniformly distributed).
If that’s a correct diagnosis, the simplest solution would be to linearly scale Permutation’s output to 0-10V.