Here's the trick how to do that: Place the standard symbol for NPN, shift-rightclick into it, change the prefix to X and the value to BFP640FESD, then copy the subcircuit into the SPICE file via the command text box. That's it. Just tried that with Q4 in your circuit -> woiks.
Basically the old Ayton-Perry trick but only half turn. Not sure if you have HoJo's book but if not, this works as illustration what you'd have without any tricks:
Take the 0402 side case but place the vias in the pads. Should get you under 0.2nH. Now place two of those, super-close to each other, run the signal trace down the middle of them (sideways), connect one on the left and one on the right. GND would then be vice versa, first right, then left. Now the loops are opposing and so are the parasitic magnetic fields, helping the cancel each other. Well, to some extent.
There are also caps with four pads but not sure how low they'll get you down in inductance.
You can also get 01005 size caps from Murata and if still not good enough parallel a few, then the parasitic inductance is miniscule. But you'll probably need lasered vias in order to have them in-pad, I don't think those can be drilled.
Have the ground plane right underneath and use a very thin board so the via lengths are small. Because they niow become a significant contributor to the parasitic inductance.
With stuff like this you either have to sit next to the layouter or do the layout yourself.
Back to the darn irrigation valve now, but this time I'll first test it in the driveway. If you hear *POP* from out west then the contraption I made to pressurize it has burst.