CCL: Do-it-yourself Freq calculations in Gaussian

 Hello members of the CCL list,
I am trying to do a Freq calculation in Gaussian03 with C140 -- two C70 fullerene cages linked by single bonds. Whereas the neutral case is very pliable and the Opt Freq combined job takes three hours, I am just having a devil of a time trying to do the same thing for the charged case -- charge = -1. I have several questions regarding this:
1). The optimization converges, but the freq calculation doesn't get beyond the first SCF Energy calculation! How is this possible since many/several SCF energy calculations were already performed during the optimization step?
2). Because my advisor was fed up with the convergence problems of the Freq calculation, he told me to calculate the Hessian myself using Gaussian and the Force keyword. So, I optimize the molecule, and then I change the coordinates one by one by delta and calculate the Forces that result on each atom. After many jobs (I use a batch file), I reassemble my results into a hessian by taking the derivative of the force with respect to displacement using a central difference derivative. I first try this with the neutral case since I already have values for the frequencies based on my successful Opt Freq run for the neutral molecule. My frequencies don't match up at all -- there are many negative eigenvalues for my Hessian despite the fact that my Hessian seems to be symmetric. Am I going about this wrong? Is there a subtlety I am missing?
3). What things can be tweaked to get the SCF energy to converge? I've already tried SCF=Tight and SCF=QC. Any other ways people know of? 
 Thank you and sorry for the long questions...
 -Connie