In the original REBO and AIREBO codes, those values are carried as twice the
actual value. The values printed in the Brenner paper should be twice the
values printed in the Stuart paper, but they use the same values in the
codes. For your case where the values are the same (case b), I think may
have been a typo in the Stuart paper. Since AIREBO was incorporated into
LAMMPS using the Stuart paper, those values may need to be changed.
Also the LAMMPS code does not handle over-coordinated carbon structures
well. I think a change needs to be made (similar to case d) to the piRC
spline when the inputs (Nij, Nji, Nconj) are too large.