TEMPO, PINT, and the Tempo2 paper (Eqn 39 of Edwards et al.) treat DM, DM1, DM2, ... as a Taylor expansion around DMEPOCH.
Tempo2 instead uses a polynomial that is not a Taylor expansion. This is noted on line 122 of dm_delays.C. For consistency with other codes, and the paper that defines the Tempo2 timing models, this should be changed to a Taylor series.