I'm not sure that I can agree with the winding schemes proposed by this utility. I understand that it is for an outrunner design, which I have no experience with, but as I understand it, the belts are not wound on individual pole pieces except in the case of 6 slots.
The belts surround several pole pieces, and sometimes have multiple layers with an overlap pattern that provides a smoothly varying phase shift of magnetism which reduces the cogging effect and makes for smoother operation.
In the illustration for 12 slots and two poles, for instance, there are two adjacent pole pieces, each of which has a clockwise winding of N turns. But in the slot between the pole pieces, there are 2N turns, but the current in them cancels, so there is zero net effect. You get the same effect by skipping that slot and winding around both poles.
But this is still very inefficient. In the slots between the poles, you have N windings at phase A, and also N windings at phase B, so the vector sum is actually 1/2 C'. It gets a bit confusing, and I don't claim to understand it fully, but if you examine commercial three phase motors (and also single phase) you will see overlapping belts.
Have you actually wound motors this way? It would be interesting to see if a different winding pattern would be more efficient. At the very least, it should reduce the overall length (and resistance) of the windings, and allow a heavier wire gauge.