I don't know diddly about the data assimilation mechanics of the individual modeling systems, but when the system was over the eastern Pacific and really only being sampled by satellite, we had pretty good agreement. (look at 00z Tue runs for example).
But as soon as the system came onshore of the west coast and presumably was sampled by the radiosonde network, the goal posts between the individual modeling systems seem to be widening. Definitely from an OP perspective and a bit from an ensemble perspective. The GEPS is way west compared to EPS and GEFS (00z suite).