In this paper, we generalise a previously-described model of the error-prone polymerase chain reaction (PCR) reaction to
conditions of arbitrarily variable amplification efficiency and initial population size. Generalisation of the model to these conditions
improves the correspondence to observed and expected behaviours of PCR, and restricts the extent to which the model may explore
sequence space for a prescribed set of parameters. Error-prone PCR in realistic reaction conditions is predicted to be less effective at
generating grossly divergent sequences than the original model. The estimate of mutation rate per cycle by sampling sequences from
an in vitro PCR experiment is correspondingly affected by the choice of model and parameters.