The heavy-hydraulics programme asks whether a small excavator does better on fuel with a stock load-sensing pump or a smaller fixed gear pump on an open-centre valve. The last two steps gave opposite answers depending on whose gear pump was assumed. Priced from Casappa's KAPPA 30 catalogue curves, a smaller fixed pump wins in 10% of 384 cylinder-geometry readings. Priced from Danfoss's GearMe Group 3 curves, it wins in 70%. The two manufacturers' curves differ by about 0.07 in overall efficiency at 150 bar.
That left a question. Is the flip a property of how efficient the pump is, or of each maker's curve shape (how efficiency varies with size, and how much the pump drags at zero pressure)? If it is the efficiency, a builder needs one number from a test bench. If it is the shape, one number would mislead.
What this step did. Both earlier steps had already swept each maker's curves scaled by a factor c, so no new simulation was needed (and this session's network refused the dataset host, so none could be run). A short script reads the committed results and places every sweep point on one absolute axis: c times the maker's efficiency at 150 bar and 2100 r/min. Five predictions about the outcome were written down before anything was computed.
What it found. Read at the pump size the fuel optimum actually picks, about 85 L/min, both makers cross the 50% line at the same place: 0.8281 for Casappa and 0.8309 for Danfoss, a gap of 0.003. Read on each family's median efficiency instead, the gap is 0.015, because the median mixes in sizes the optimum never chooses. Giving Danfoss's curves Casappa's idle drag widens the gap instead of closing it, so drag is not what separated them. Three of the five predictions held. One that failed matters: above the flip the two sweeps part (59% against 76% at 0.84), so once a smaller fixed pump wins, how often it wins still depends on more than one number.
For a builder: the question has narrowed to one bench test. Does the fixed gear pump you would fit, about 40 cm³/r, reach about 0.83 overall efficiency at 150 bar and working speed? Read at that size, Casappa's catalogue curves say about 0.76 and Danfoss's about 0.85. An ISO 4409 test report would settle which is typical.
Script, output and figure: experiments/gear_pump_crux.py, gear_pump_crux_results.txt, gear_pump_crux.svg. The predictions and their grades are in the project's LOG.md. Both makers' curves are published "typical" selection data, not test reports, and nothing here was run on a real machine.