Correct CVT infinitely variable within hardware constraints (includes MG1 and MG2 rpm constraints and the final drive gearing. The best one could do (tallest overall gearing) is to put a holding torque on MG1 so that it does not spin when the engine is running. Then, all engine rpms would be directed to the output shaft. So, a numerically lower final drive ratio would yield lower engine rpm than a higher final drive ratio at the same vehicle speed. For highway cruising one wants to run ICE at the most efficient area of the Brake Specific Fuel Consumption map which is generally low rpm, high torque. It's like up-shifting a conventional transmission to a higher overdrive gear to lower engine rpm to improve FE. This should be achieved in Negative-Split Mode Operation (from the Ford OBD System Operation Summary for Plug-in and Hybrid Vehicles):
Prior to a PCM update of the Hybrid to increase the EV top speed from 62 mph to 85 mph, it was quite easy to get into negative split mode by controlling the throttle and to stay in this mode for some time. With EV operations now up to 85 mph, it harder to get into the mode (keeping the battery charged near its upper limit) as any slight decrease in power requirements can trigger EV operation at higher speeds.
The different final drive ratio for the hybrid and NRG supposedly was to improve performance of the NRG given it's extra weight (larger HVB) such that both vehicles would be very similar performance wise. My guess is the PCM algorithms for both the NRG and Hybrid are the same once the NRG enters Hybrid mode operation. So, by numerically increasing the NRG final drive over the Hybrid ratio, the NRGs low end performance would be improved somewhat.
From what I can see the Negative Split Mode is bleeding off excess energy because the battery is full? It says "the generator consumes electrical energy". Also, they list a "parallel mode", but don't describe it with diagrams and an explanation such as they do for the Split, Series, and Electric modes.
I notice that the document diagrams of the Negative and Positive Split modes are identical from what I can tell.