Question about branching plane updating algorithm

Dear developers and forum users,

I have read the references [JCTC, 2010, 1538] and [JPCB, 2014, 7806] in manual. But I am not sure whether I understood correctly. So I carried out a test job of MECP optimization on a simple molecule.

I calculated gradient as eq.10-11 in [JPCB, 2014, 7806] from the outputted energy gap, mean gradient, vectors x and y. However, my result was not the same as “Full analytical gradient” in the output file.

Beside, ref [JCTC, 2010, 1538] mentioned that mean gradient is used as the initial vector y0 for updating BP. But I found it was not such case in Q-Chem output (y0 \neq Gmean). It seems that Q-Chem projects x out from y and then normalizes it to yield y0.

My question is that how does Q-Chem calculate force/gradient for MECP optimization using BP updating algorithm.

Thank you for your help!