The role of virtual simulations in the evolution of primary school students' mental models about electrostatics is investigated by analysing the answers of two groups of 9-10 years old pupils inside instructional sequences that combine two hands-on activities using balloons and jackets with the same activities using PhET simulations: one group followed a sequence where the real experiments were conducted before the virtual simulations, the other group followed a sequence where the virtual simulations where proposed before the real experiments. Students' answers were categorized into a descriptive dimension (accordance with the phenomena) and into an explanatory dimension (level of adequacy with the microscopic model based on elementary charges). Our results show that, although virtual simulations improve the level of adequacy of students' answers to the target model, this improvement is not transferred to new phenomena, so that the capability of relating the developed model to the real world is the same in both sequences. Moreover, about 40% of pupils remained at their starting non-explanatory level in both teaching sequences. We conclude that simulations per se do not help students in evaluating and refining their models, indicating the need for teachers' support in modelling-based instruction involving not visible physical entities.