An investigation is carried out looking at correlations between the orbital elements of collisional targets and projectiles, estimating the number of interlopers in Trojan collisional families, and at the possibility of determining the ages of the Jupiter Trojan families by orbital integration. Real Trojans are integrated and close encounters are recorded in order to evaluate collisional circumstances between Trojans. Fictitious collisional families are created and integrated for 10 MJyr (million Julian years) forward in time and back again to the time of the collision in order to check the performance of the integrator, and the behaviour of the fictitious collisional fragments. Proper elements are calculated for the detection of family clustering using the hierarchically clustering method. This method presents little difficulty finding fictitious families in the Trojan swarms even in areas with densely populated backgrounds. However, even when the background is relatively sparse in objects, several interlopers can be connected to the family at velocity differences below 100 m s −1 . On the other hand, in densely populated backgrounds the contamination of interlopers should be less than 30 %. Providing gravity is the only significant force acting on the Trojans and resonance effects are weak, the shape the collision fragments create in the proper element space are preserved on the GJyr scale, and collisions can be tracked with orbital integrations for ages of at least 100 MJyr. However, the shape of artificial families does not correspond to suggested real families. This points to the need of including non-gravitational forces such as the Yarkovsky effect in order to simulate the family evolution. As a consequence age determination by orbital integration might be severely restricted and previous investigations involving long term orbital integrations might have to be recalculated.