The reliability of a system depends on the joint effects of ageing and maintenance efficiency. Its assessment requires the application of probabilistic models and the statistical estimation of their parameters from the observation of field data. The considered imperfect maintenance models are the virtual age models of type ARA (Arithmetic Reduction of Age). We are interested here in the case where the data are interval censored: we know that a failure occurred between two preventive maintenance dates, but we do not know exactly when. To estimate the parameters of virtual age models in this situation, this article considers two different hypotheses on the effect of the corrective maintenance following failures whose dates are censored. In both cases, pseudo failure dates are generated to replace the censored dates. The study compares on simulated data the loss of quality of estimates when using the fictitious dates instead of the real dates.