Hi Torkrench,
Do check with a scope the output of the audio oscillator, when running in internal mode.
On the collector of Q104 you should see sinusoidal signal, about 800Hz, about 8Vpp, and not that much distortion, about 3% - 4%. (This is from simulation.)
The internal modulation changes "modulates" the supply voltage of the RF-oscillator circuit.
It is done by the emitter follower, Q103, which works as a pass transistor. At the emitter you should see similarly "undistorted" sinusoidal signal, about 3.3Vpp, "sitting" on the supply voltage.
The sinusoidal amplitude on the supply voltage depends on the "raw" DC supply voltage, measurable on the collector of Q103, that I estimated 16V.
The bottom line is that you should see that "undistorted" sinusoidal signal on the supply voltage.
If you see that, then the rest is done by the RF-circiut in the generator, meaning that possibly it is as good as it gets.
"Plugging" the simulations together (the audio and RF-stages) this is what you get.
And here is the FFT of the output. The RF fundamental frequency is 2MHz.
Regards, Peter