EFA attempts to explain the correlations not the variance. I would decide on the factors based on model fit statistics and substantive reasoning. The reason to select a model with five factors rather than four should be motivated by the number of factors expected from the way the items were constructed and the meaningfulness of the factors. Factors should not be extracted for the purpose of explaining variance.
thanks a lot for your opinion. I'm glad you're arguing from a logical rather than from a statistical point of view. The four factors that were originally extracted make perfect sense and I'm happy to have found such meaningful and distinct factors.
I was just worried that somebody could argue that those factors do not explain enough of the overall variance of all my variables.
So instead of the overall variance, would you look at criteria such as Bartlett's test and KMO?