In a recent paper, Loveday et al. presented new results on the luminosity function for a sample of galaxies with b(J) less-than-or-equal-to 17.15. After having morphologically classified each galaxy (early-type, late-type, merged or uncertain), they estimated the parameters of a Schechter luminosity function for early- and late-type galaxies. There is, however, in their sample a bias against identifying early-type galaxies at large distances and/or faint magnitudes: in fact, many of the early-type galaxies at faint magnitudes have probably been classified as 'uncertain'. As discussed in Loveday et al., the existence of such a bias is indicated by the fact that, for these galaxies, [V/V(max)] = 0.32. In this paper, we show, both theoretically and through the use of simulated samples, that this incompleteness strongly biases the derived parameters of the luminosity function for early-type galaxies. If no correction for such incompleteness is applied to the data (as in Loveday et al.), one obtains a flatter slope alpha and a brighter M*, with respect to the real parameters.