Martin: Did you check that your nipples, eyelets and spokes are not ferromagnetic? These can deliver power relatively uniformly on account of the ferromagnetic effect.
Otherwise, I continued experimentation producing more numbers and the main is the partition: When using a Magnic type arrangement, with a bicycle wheel that lacks ferromagnetic components, except for the steel insert closing the rim, of the order of 58% of power from the Magnic device is produced by the ferromagnetic effect and about 42% by Eddy currents. The error in my measurements can be easily 5-10%.
I found it hard to experiment directly with the rim, because of the inability to suppress the operation of one of the effects, ferromagnetic or Eddy. A round radially magnetized magnet would eliminate the ferromagnetic effect, but commercially you cannot get strong radially magnetized magnets, so you need to arrange individual magnets along a radius in a circle, i.e. e.g. N out, as I suggested to Martin. In practice this, however, does not work, because the resulting field of the manets is not uniform in angle and the ferromagnetic effect remains sizable in spite of the effort.
With this, I gave up experimenting with the rim, and took instead an alu plate, about 2 mm thick and mounted it on a drill. I mounted six N42 cube magnets 9mm on the side, onto a brass circle and arranged them with the alternating poles out, as in the Martin experimentation or in the Magnic patent. I placed my rotor under the rotating alu plate, either with or without a steel nut glued on top of that plate. The axle of the rotor was connected to a bobbin to which a string with a hook was tied. On the hook I hanged washers to test how much torque I could get from the rotor. The drill driving the alu plate was operated at 1720 RPM and I placed the rotor under a location about 27 mm for the center of the rotating plate. That gave the equivalent bike speed of 1720*60*2*pi*27*10^-6=17.5 km/h. The nut that was glued to the top of the plate was chosen so that it would occupy the same fraction of the circumference as the steel insert in the rim. The washers that I could hang on the string, so that they would be lifted or at least suspended in the air, without or with the nut glued to the top of the plate, are shown in one of the linked photos. The weight of the washers + hook + string was measured in the two cases to be 5g and 12g. In the first case only the Eddy effect operated and in the second - both Eddy and ferromagnetic. The weights tested the torques and correspondingly the powers that this or similar rotor could deliver without and with the ferromagnetic effect. The ratio 5/12 yields 42%.
To complete the measurements I tried to have a nut attached to a material that did not provide a conducting path for Eddy currents. With the ferromagnetic effect alone, however, the operation of the rotor was too jerky to make proper measurements. I increased the drill speed twofold and then the ferromagnetic effect alone would allow to lift the same washer as with the Eddy and ferromagnetic effects together at half the drill's speed. Still the operation was quite jerky and unreliable, so I could not really tell more in that measurement that the effectiveness of the ferromagnetic effect due to an insert was not inconsistent with being about as strong as Eddy.
One test that I tried was to put my rotor on top of a regular bottle dynamo. There was no way to activate this way the dynamo at regular bicycle speeds. One can say that the emergence of a Magnic type dynamo is completely due to the LED revolution and their low power requirements. The LEDs open new options for power delivery.
Above my numbers, when there are steel spokes or nipples, the ferromagnetic effect should overwhelm Eddy.
Click here for photos.