Comparing two Rohloff bikes is almost going off-topic except it highlights the difference of making comparisons in other than laboratory conditions.
Yes this. AJ's testing has so many variables that isolating one element isn't possible, it becomes anecdote not data.
That may be comforting but it is nonsense, as I shall explain shortly.
Martin has the right idea:
Horses for courses. Just how do you measure efficiency?
Not very scientific, but I just ride the bike several times over the same circuit before and after changing the gear system (or other component) and compare the average time taken.
On the contrary, perfectly scientific, the more tests you take, the more scientific.
All the four bikes I cited went over the same 7.5m circuit in the same direction more than 440 times, year-round. Thus every bike partook of all the offered weather conditions, all of my conditions that may influence speed, etc. Two of those bikes were Dutch vakansiefietse with Shimano Nexus boxes, the only difference being that one had a fully automatic gearbox and adaptive suspension aiding efficiency; if any upset in the results were likely, it would have shown in these two. The other two, derailleur and Rohloff, were steel, another layer of indicators in case there was something wrong with the count. I wrote down the order I expected before extracting the data, as a hypothesis, if you like, and the result was precisely as I expected. You're arguing against a huge, huge statistical probability for my result being rock-solidly unimpeachable.
1760-plus rides over the same piece of road on four bikes is a long way up the ladder of certainty from an anecdote, laboratory or none. That large a sample smoothes out the small daily fluctuations until hardly a ripple disturbs the result.
You're the one trying to hold up anecdotes of single rides as contrary evidence, Paul, as in your post above, not me.
***
In any event, by the evidence of the net, the average Rohloff tourer isn't a kid wet behind ears and possibly a fashion victim but a mature party of judgement and means. He can, if he wishes, scour the net for Rohloff information, and form his own judgement; if he is serious he is likely sooner or later to land either here or, if he speaks German or has discovered Google Translator, on the good German board, where the various opinions eventually combine to form a singular opinion, which is then his initial opinion. We're just privileged to be on board not too far after the beginning of Moronic's journey. We won't mislead anyone by a variety of pretty much mainstream opinions.
As for a laboratory, back when I took these rides, I financed a prototype lab to build and test the high tension (up to 2000V) thermionic tube (valve) audio amplifiers I designed for a Japanese maker. My engineering partner had spent his life at the cutting edge of instrumentation. He was no faculty common room talker: he'd done it daily all his life. When I suggested to him that we should test my new Rohloff, he pointed to the tarmac on the parking lot outside and said, "The best test is the one you already performed on your earlier bikes. Match it." So that's what I did. When I caught up, a bit late, on Kyle and Berto's Rohloff v. derailleur tests, I saw instantly what he meant: those tests were pie in the sky, useless for real life decisions. A one or a two per cent efficiency difference when the power source is outside the equipment being measured, linked by a chain, will just disappear into any number of variables, like the fact that their Rohloff wasn't run in but brand new. Ouch!
And I repeat: in real life the Rohloff's efficiency is the average over the ride, which is very likely to be a constant over many rides on the same course, whereas the derailleur's efficiency is not fixed when it leaves but should be measured and calculated upon its return, dirty and degraded. The Rohloff average efficiency will be the next best thing to a flat line, the derailleur efficiency line will always, except for a brief honeymoon early in its life, point downwards.
By the way, Bernd Rohloff's reply
https://www.hupi.org/HParchive/PDF/hp55/hp55p11-15.pdfto Kyle & Berto gives a whole bunch of considerations, including ones from sports medicine that I picked up on immediately I read the K&B article, of the type that my own tests would subsume under straightforward statistical weight of number of tests.
But for my money the best thing that Rohloff and Greb do in their reply is "Figure 6: SPEEDHUB 500/14 efficiency comparison. Gears 8-14 shifted to the left to compare with gears 1-7." It shows why I'm so keen on Gear 11 that my whole transmission is planned around it and the overdrive gears come into use here in the Rome of West Cork (very hilly!) only on the downhills.