Good and normal values. We can forget the theory that Q5 is feeded bypassing the LM10 through R37 - not with 3.9M.
No other suspects.
So, let's go to the basics.
First: just for sure: The 300V output is bufferd by C7. Am I right that you do not have any oscilloscope equipment to check the 300V for any ripple? If not, try to measure the 300V ouput with a DMM in pure AC mode. You should not get a value of more than a few volts. If you have a ripple of 20-50v, there's maybe an issue with C7, C17 should buffer the (by Zeners) downscaled 300v output peaks, so the oscillator is regulated not to exceed 3xx volts. But with a significant ripple you will meausre a lower DC average on your DMM, and the output lowers more with any load than usual.
Note: When measuring in AC mode ... If you have an idea which frequency the oscillator works ... this could lead to false interpretations of AC readings at this point. I just checked my DMMs for that and got weired results with older (non rms) DMM, whereas two new OWON HDS120 and HDS242s are blind with frequencies >10 kHz. If you have any sine oscillator (AWG etc), check your meter for accuracy with various frequencys (if not done already, it's always good to know).
Second: The Zener voltages are 91v, 92v and something between 88 and 91.5v, depending on the R34 position. Given you measured while R34 was at the low end (and C17 voltage 3.9v) all three Zeners are equal, otherwise there may be a problem with CR13. But I do not think so.
Third: Did you find any output voltage variation when turning R34 from 0 to 1k, and if, what variations of Q5 base or emitter voltage you see? If you find the output variation of let's say 20v and 0.X v at Q5, this may be normal.
Fourth: If you can vary the output with R34 and your are shure that C7 is ok and there's no significant ripple on the 300v output (a few volts are ok), and only then!, try to lower R28 a bit with a bypass resistor (start with ~ 5MegOhm in parallel). 3.9 volts on 100k result in a 0.039 mA current through them, resulting in ~12 mW total power dissipating, if I do not fuck the maths. With a 1MOhm in parallel, current and power dissipating rises by 10%, should be safe. 0.034 mA (or 4 mW) are not much for a Zener (rather 100mW are), and in this area they have a rather soft U/I characteristics, with only a bit more current you should get up to 100 volts per Zener and thus more output voltage. Do we have a spec for the Zeners? I do not think they are 91volts, rather 100 oder 110 volts.
Start the circuit with R34 on the 1k side and try to rise the output voltage.
Good luck!