WOWJust had a 3 hour conversation with erfinder on skype, we just worked out what is going on in your armature. Its coil shorting and it is what is causing the gain in voltage in the output coil. Its very complex and it will be difficult to explain via posts on here, but the basis is this.
A lap wound armature is effectively a loop of coils in a ring. As the commutator moves under the brushes it effectively shorts out one of these coils (actually 2, one under each brush). This has 3 effects
1) The shorted coil has capacitance and so tries to discharge the stored energy across the short, as the resistance is very low a huge current flows in this coil and inductively transfers the energy to our generator coil.
2) During the coil short, it changes the inductance of the whole armature lowering the inductance allowing more current from the supply to pass through the windings charging them up.
3) As the shorted coil opens again the inductive kickback jumps in voltage to pass its remaining current into the whole armature, remember the inductor wants to maintain the same current regardless of Impedance so the voltage jumps up way above the source so that this can happen. At exactly the same time the inductance of the whole armature jumps up causing another inductive kickback trying to maintain the same current so the voltage jumps higher.
The net result is very high voltage in the armature which causes arcing.
This energy is normally dissipated in a compensation coil but because we have a load (Resistance) on our generator coil this does not happen efficiently.
This is why there is a second set of brushes on the armature, to collect this charge before the armature arcs. This charge is then placed in a capacitor across the source line reducing the input.
The key thing here is that the inductors in the armature are charged under a condition where their inductance is low and discharged when the inductance is high.
It is another gain in the systemThe reason why we have such a high current drain and low efficiency is because:-
a) We have lowered BEMF due to the power coils position, causing a higher current drain.
b) We are pulsing at the frequency of 60hz and not at the frequency of the switching (coil shorting) and not only when the inductance is low therefore lowering efficiency and pumping high voltage into your variac.
To solve this problem we have to commutate the pulses so that they occur at the right time and for the right duration. This explains why the brushes have to be only one commutator segment width for the power going to the armature, to some extent. We need to pulse the motor at the exact time that the short begins and stop the pulse at the exact time the commutator breaks the short. the rest of the time the armature is powered by the inductive kickback.
Man this is complex. This is assuming that the lockridge did indeed work this way and not how I originally thought. If it did not work this way then myself and erfinder have just come up with another way of doing it

For now, we will continue with my original plan as it does not stop us from progressing this new theory, but we need to try and recover this high voltage out of the armature with another set of brushes. I will put some more thought into how we are going to pulse it.
The picture below was created during the discussion, If you want me to go through it with you it would be better to do it on skype.