After thinking about this and doodling on paper, I believe the flaw in my thinking is that because the pushrod cup and the rocker roller are not on the same "plane" (maybe not the right term, but hopefully it will still be understandable) w/one another, the roller will describe the motion as depicted in the diagram, and Bogie's description. If the two points were perpendicular(?) to the rocker stud, or if the roller were lower than the p-rod cup, my theory might hold true. But because the roller is positioned above the p-rod cup, the movement is as Comp/Bogie said.
I do believe that if the lift were high enough that there's a possibility that an 'ordinary' rocker arm might go towards the exhaust side per Comp/Bogie, then at some point actually reverse and move back towards the intake side again. But in practice I doubt this will occur until lifts were enormous and even then the design of the rocker might be able to counter that effect.