The average-WHR based balance generally does reasonably well but there are corner cases where it's probably not great.
For instance, it seems pretty likely to me that if three players
of approximately equal skill play a 1v2 the arrangement likely to produce the most interesting games is if the (marginally) strongest player plays alone against the other two.
Certain kinds of parties/clan arrangments can tie the balancer in knots also.
These corner cases do not appear all that often so it is difficult to make a statistically meaningful analysis of them or decide what the best thing to do is when they arise. I am not sure that a specific attempt to fathom the corner cases has been made. Generalising about them from non-corner cases is likely to work out quite poorly.
(The quality of balance in some weird configurations is also affected by the map, which the balancer doesn't take into account at all. I expect 1v2 is way less winnable on maps where the enemy team can split your attention, starting with your coms in the same place is a crippling disadvantage, etc.)