Another set of random thoughts:
Looking at the diagram above, there is a lot of leverage between the cross shafts and the spindles. Any play would cause a problem. Also, are the spindles fixed into the car securely - i.e. is the play from spindle to housing or from housing to car?
I think the old (trad) Morgan wiper rack had a lot less leverage between actuator rack and spindle.
I would not be surprised if, when you conduct the arm removal experiment to see if one blade is worse than any other, that you find it to be one of the blades nearest the motor. The arrangement shown will have forces flying in all directions if there is any slack.
Hoping this is helpful.
