It is well known that
volcanic forcing appears to have an efficacy materially below one, at least when used in simple climate models: see the discussion in Lewis and Curry 2014.
Not exact matches
The main difference between projections A and B
appears to have been the projected
volcanic forcings in ~ 1995 and ~ 2015.
The response of the models to a true
forcing of
volcanic eruption
appears to be correct.
The OP is making an assumption that
volcanic eruptions apply a
force to planetary temperature, which is then free to do as it likes within the reference frame of the planetary climate system and that
appears to be causing it to flip back into place.
Although Marvel et al. do not mention the very low efficacy of solar
forcing in their simulations, this
appears to have more effect on ERF efficacy for the sum of
forcings over the historical period than does low
volcanic efficacy.