The harmonic shift implemented in
#591 is missing the case discrimination for the conformalMethod parameter; the way it is currently coded is correct for conformalMethod = 1 (W method) but not for conformalMethod = 0 (phi method). My original version of the code had the two variants but I removed this at some point while testing and my submitted patch contained only conformalMethod = 1. In my original notes, I have this
alpha^2 em4phi (gtu[ua, uk] (PD[alpha, lk]/alpha + 2 IfThen[conformalMethod, -1/(2 phi), 1] PD[phi, lk]) - gtu[ua, ul] gtu[uj, um] PD[gt[ll, lm], lj])
which should be checked again before committing.