I think that DerS. has a point about the primary in relation to the secondary. I have always had them about 1/4 of the way to the top, and supported 1" off of the secondary windings. I never tried it completely below the secondary, but I don't think you will get optimum electromagnetic coupling that way.
I have gone through many sets of FETs and drive chips with this type of design. Even with the best shielded case that I can make, I still would blow up the FET stage. I am almost certain that it has to do with coupling of the radiated signals back into the drive circuitry, but it happens fast and pretty dramatically. Usually I have it running for a half a minute or more without problems. And then only when I interact with the generated arc, then things go bad.
Driving directly from 110 VAC, with a similar push pull design, we get well over 1" of arc. But the point to point one is the only one that consistantly works. Using the same design, we built both vertical and horizontal units. Two of the horizontal are running fine, 3 versions of the vertical keep failing. The next for solving this is using two sided PC boards with a solid ground plane on one side to reduce the couple noise into the FET drive circuit.
Good luck,
JimW