Dennis,
As Bart mentioned, JavaTC doesn't take into account spark loading.
Capacitive spark loading can make the actual tune point somewhat different than
the JavaTC tune position. As Bart mentioned, he's seen up to a 13% difference
in tune point due to spark loading. I think it would be interesting if Bart added another
additional tune position output for his program which would suggest a tune point based on
spark loading, based on the expected spark length for a given input power and
break rate, etc. This could get tricky though. It would just give a rough tune point,
but less rough than not accounting for spark loading.
Anyhow based on all this, my guess is that you have fewer than 1650 turns, and that
you do indeed have heavy build insulation on your wire. It's the spark loading issue
that's confounding everything. If you do a scope resonance test using a signal
generator, it will give the secondary frequency without spark loading, and should
show a higher frequency (fewer secondary turns). Another test you can do drape
a breakout rod over the toroid, extending 6" over the side of the toroid, and run
the coil at very low power (barely breaking out), and retune the primary for longest
sparks under that condition (with very short sparks due to low input power). This
way spark loading won't be much of an issue, and you'll get a truer tune point
based on number of turns on the secondary, without being confounded as much by the
spark loading effect.
Cheers,
John