In the tokamak JET it is observed a systematic discrepancy when comparing the neutron yields as measured by KN1 diagnostic (fission chamber) with the results of codes assuming no fast particle transport beyond collisional diffusion (for example PENCIL, CHEAP, Nubeam, ASCOT). As data values from simulations are typically higher than the experimental ones, this is called 'neutron deficit'. On the other hand, recent studies suggest that one of the mechanism through which the NTM island degrades tokamak performances is the expulsion of part of the fast ions from the core of the plasma, thus reducing the rate of nuclear reactions, and ultimately producing a neutron yield lower than expected. Fast ion dynamics simulations with the Monte Carlo code ASCOT, modified to account for the magnetic island presence, confirm qualitatively this pattern when compared with tokamak data. In this work we study how the comparison between experimental and simulated neutron yields changes when the island effects are implemented into a transport code. The suite of transport codes JINTRAC has been specifically modified and the island parameters are determined from experimental data.