Improving a found TS with ReaxFF

Dear LAMMPS experts,

I have observed a reaction in my simulation using ReaxFF as a force field. To characterize the reaction using ReaxFF, I ran NEB calculations using 100 replicas all of which were obtained from a reactive trajectory. The best NEB result I got for TS (climbing replica in the log.lammps file) has a gradient of ~2 kcal/mol.A that is not very satisfactory.

I am thinking of relaxing this structure so that its gradient improves. However, using minimization the inter-atomic distances change as well which leads the structure to either reactants and products. I wonder if there is any way in Lammps to do so?

In general, is it possible to get the TS of a reaction predicted by ReaxFF, using a guess that comes from the NEB method?

Thank you so much for helping me.