Hey, I know that Schall guy pretty well 
You might want to first check your density for your starting coordinate system. I quickly calculated it based on your number of atoms created and the sphere radius and got a number quite a bit higher than that of bulk diamond.
vol = 4/3pi * 10ang^3 = 4187ang^3
dens = 1658 atoms * 12.01 g/mol / 6.02e23 atoms/mol * 1e10^3Ang^3/m^3 / 100^3 cm^3 / 4187Ang^3 = 7.9 g/cm^3.
That will give you all kinds of grief and trouble, trouble you may be covering up during the initial stages using temp/rescale. For DLC shoot for a density between 2-3 g/cm^3, bulk diamond being 3.5 g/cm^3. See if that helps. Also get rid of the temp/rescale and use Langevin or Berendsen for your thermostat and see if you get stable dynamics during your initial stages. You might also try dropping your time step down to about 0.25 fs as well.
Good luck. Your idea to use the indent function to get a spherical tip is pretty clever. I might just steal it 
Dave Schall