DMol3 partial Hessian questions

I have been trying to use partial Hessian calculations in DMol3 (v. 4.2) and have run into the following problems:

1) One of my Hessian calculation was aborted (exceeded walltime limit) after all displacements were computed, but before the frequency analysis was finished. I tried to restart the calculation from the .hesswk file. The output I get is this:

HESSWK file is complete, continuing with analysis

Then it gives a warning for each movable atom:

Warning: Hessian data will be augmented for atom # NNN

and proceeds to recompute the .hesswk file from the start.

Why does it try to recompute the .hesswk file, and how can I avoid it?

2) When I visualize the vibrational modes obtained with a partial Hessian, it shows moving atoms which were excluded from the Hessian. They are also very active in modes with imaginary frequencies, while positions of these atoms were previously optimized to a minimum. This creates a problem for my transition state optimization. I cannot afford to use the complete Hessian due to the molecule size. How can this behaviour be explained, and what tools should I use for the TS optimization? The reaction in question occurs in a zeolite, and I had all of the zeolite frozen, except for the binding site.

I would appreciate any advice

Vladimir Shapovalov