As with many of these types of motors and mechanisms, trying to understand exactly how they produce energy can be confusing without building one. I will have to agree with Aaron here as I have had some experience with this device.
When a Milkovic type oscillator is first assembled and operated, it behaves much like the theoretical model would be expected to behave. The output is erratic and weak, the pendulum is difficult to keep going and the whole thing is an inefficient mess.
However, it’s a whole different machine once it falls into resonance. The output becomes strong and regular, the pendulum takes very little energy to keep it swinging and you can see immediately what Milkovic is talking about.
You really need to build one and test it to get a better understanding of the principal involved. Then you can try and get it into resonance, which will teach you in no uncertain way the difference between a resonant and a non-resonant device.

Ted