As I am running all my test with a regulated laboratory power supply, I not only can control the voltage, but also the current of the power source. Having done that, I realized that more power shows up at the output of the lowpass-filter at a particular reduced current. I measured the input current of the 74HCT00 and found out that it best performs at about 15-16mA (??, but OK). A little math in my head, and I added a 220Ohms resistor between the regulator and pin 14 of the quad-nand.
But now, running the NANDs in parallel, with optimum waveform and output, made the frequency shift much larger (???), something like 20Hz, in place of 5Hz before (!). Additionally, the frequency slightely dropped (well, ok, that I do understand).
No math helps here, pure experimentation. Remember, the shift is predominantly determined by the resistor connecting the HCT00 and the 12F629...
For sentimental reasons, I left in the 47kOhms resistor which started it all. It has now got a partner on its' side, they are even allowed to hold hands, a 470kOhms resistor in series. Seems that in the new arrangement, 0.5MOhm makes the trick, shift is now back to 5Hz. Ouuffff!!!
Seems there is a delicate balance between the few critical elements. And now limiting the current seems a factor too.... Am I using the HCT-gates in some sort of analogue mode?
And what about the results? Well, here we are. @ 50Ohms, I measured 0.8Vpp, resulting in about 1.6mW. Actually, the waveform looks even better now! Presently running a stability test (RX=FT817 w/TCXO).