Hi Ben,
I'm not sure I follow that exactly, but I'm thinking it might have to do with only the endpoints being Geocorrected and things assumed to be linear in between. That's the only difference I can think of between Jogging and GCode Motions. When you move with GCode/MDI are you looking for errors during the motion or at the endpoints of motion? See diagram below:
Motions will be automatically sub-divided into segments less than an allowed maximum length of:
MaxLength = FeedRateToUse * MP->TPLookahead/2.0;
What is your Trajectory Planner Lookahead time set at? What Feedrate are you running?
Rapids will not be subdivided unless the "Rapids as Feeds" option is selected.