Agreed on design considerations. I had no intention of operating anywhere near stall torque. I just knew from experimentation I needed X torque to "barely turn", so I just figured to buy something whose stall torque was roughly 3-4 times that. It was only then that I really scrutinized those charts and started the confusion.
Since my last post I googled gear train efficiency for various gear types. References disagree, but they all fell into the range of 92-98% for spur gears. Taking the most pessimistic view (92%) it would require an additional 5 gear stages (0.92^5) to drop power output down to 65% of input power. So, something is fishy. We're either missing some piece of the puzzle, or those numbers were simply generated with far more pessimistic constants.
At this point I think experimentation is the only way to know.