Radiative feedback is among the most important consequences of clustered star formation inside molecular clouds. At the onset of star formation, radiation from massive stars heats the surrounding gas, which suppresses the formation of many low-mass stars. When simulating pre-main-sequence stars, their stellar properties must be defined by a pre-stellar model. Different approaches to pre-stellar modelling may yield quantitatively different results. In this paper, we compare two existing pre-stellar models under identical initial conditions to gauge whether the choice of model has any significant effects on the final population of stars. The first model treats stellar radii and luminosities with a zero-age main-sequence (ZAMS) model, while separately estimating the accretion luminosity by interpolating to published pre-stellar tracks. The second, more accurate pre-stellar model self-consistently evolves the radius and luminosity of each star under highly variable accretion conditions. Each is coupled to a raytracing-based radiative feedback code that also treats ionization. The impact of the self-consistent model is less ionizing radiation and less heating during the early stages of star formation. This may affect final mass distributions. We noted a peak stellar mass reduced by 8 per cent from 47.3 to 43.5 M⊙ in the evolutionary model, relative to the track-fit model. Also, the difference in mass between the two largest stars in each case is reduced from 14 to 7.5 M⊙. The H ii regions produced by these massive stars were also seen to flicker on time-scales down to the limit imposed by our time-step (<560 yr), rapidly changing in size and shape, confirming previous cluster simulations using ZAMS-based estimates for pre-stellar ionizing flux.