The simulation result depens a lot on the used tube model.
But below is one I have used and the actual performance is very close to simulated one.
The grid resistors will be smaller depending of tubes selected unless a williamson type driver tube is used.Yes fine, but only way to extract full performance is to feed into Williamson diff drivers or straight into a low Pout class A op/stage; Morgan Jones did the working out for the Mullard 5-20 and found the long tailed quite heavily loaded esp when the op/stage grid leak resistor values is reduced, then it starts to produce more thd and also by Miller effect. Fussy with distortion ? Look at the big picture with the whole design; wrap it up with copious amounts of global NFB. That's exactly what many early designs did.
The grid resistors will be smaller depending of tubes selected unless a williamson type driver tube is used.
With 6550 or kt120 it should be 50k.