I don't think the problem is in the super-regen receiver per se. That was all we had for years. I still suspect the problem is in the coding/decoding. Either the transmitter or receiver isn't handling it. When we had signle-channel, super-regen receivers and escapements, our pulse rates were manually generated and were very low, because the feedback loop was what the airplane did in flight. Albeit, helicopters are a lot more sensitive than our intermittantly guided free-flight models were, but for motor power, I think you can still go lower in frequency.
Now, why did the battery charge have an effect? Do you have effective capacitors and/or other noise suppression on your motor? It does seem odd, though, that the receiver can decode reliably the transmitter signal, but not simple on/off modulation.
As for low frequency PWN, obviously you can do it with discrete elements as you have shown. Last week, however, I bought a little signal generator on e-bay for about $120 USD that goes to fractional hertz, has adjustable duty cycle, and display of frequency and amplitude. I don't know if yout tinkering budget would include something like that, but I am very happy with it. The brand in the USA is Victor VC2002. Here's a link:
**broken link removed**
John