Vehicle in the example I gave is a 2013 Dodge Grand Caravan. If total wheel slip is detected on one side, antilock is applied to both sides. Otherwise the van would tend to go into a skid and spin because one side is braking and the other is not. Since anti-lock pulsing occurs on all tires if one track happens to be on polished ice, which has basically no friction at all, the result is esentially no braking whatsover. The traction that is available on the other 2 tires on dry road is not utilized. Antilock seems to be designed with the primary function of keeping the vehicle going straight, sometimes at the expense of quicker stopping. If I could turn off ABS for winter driving, (at least in that vehicle), I probably would. Since learning to drive as teenagers, we had plenty of (intentional) experience sliding cars on snow and ice, and keeping them under control.I have to think that studies like the one referenced in the OP account for confounding factors like you describe. In the case of two wheels on ice and two on dry pavement and a multiplexing 2 or 4 channel ABS the two wheels with good traction would brake to the point just short of lock up wouldn't they? The amount of braking pressure applied by the leg and foot wouldn't match the deceleration rate and it might feel like the van isn't braking at all when it actually is.
It seems to work this way with ABS motorcycles I've owned. If the rear brake is over applied all you feel is the foot pedal tremble or vibration but the bike is braking and decelerating rapidly due to weight transfer to the front tire but when the front wheel ABS kicks in you do feel like no brakes are applied at all.
Of course, none of this applies to motorcycles since they are single track vehicles. They only reason I brought it up was in response to TacomaJD's post.