This problem has been resolved already. The inconsistency between lammps and reaxff is because you need to manually provide seed to generate initial velocity in lammps, for temperature of 3000K, a line like this is needed:
velocity all create 3000.0 54912115 dist gaussian sum no rot yes units box
Check the lammps manual for velocity to find more.
Fix reax/c/species determines bonds and molecules based on
user-defined bond order cutoff values, instead of cutoff distances.
Therefore for dense systems if ReaxFF is still producing the correct
connectivity, hence the correct bond order, there should not be a
problem for fix reax/c/species.
On the other hand, BO values may increase for dense systems due to
shorter bond lengths. It may help to define slightly larger BO cutoff
values in fix reax/c/species for denser systems.