(Abridged)The X-ray measurements of the ICM metallicity are becoming more frequent due to the availability of powerful X-ray telescope with excellent spatial and spectral resolutions. The information which can be extracted from the measurements of the alpha-elements, like Oxygen, Magnesium and Silicon with respect to the Iron abundance is extremely important to better understand the stellar formation and its evolutionary history. In this paper we investigate possible source of bias connected to the plasma physics when recovering metal abundances from X-ray spectra. To do this we analyze 6 simulated galaxy clusters processed through the new version of our X-ray MAp Simulator, which allows to create mock XMM-Newton EPIC MOS1 and MOS2 observations. By comparing the spectroscopic results to the input values we find that: i) Fe is recovered with high accuracy for both hot (T>3 keV) and cold (T<2 keV) systems; at intermediate temperatures, however, we find a systematic overestimate which depends on the number counts; ii) O is well recovered in cold clusters, while in hot systems its measure may overestimate by a factor up to 2-3; iii) Being a weak line, the measurement of Mg is always difficult; despite of this, for cold systems (T<2 keV) we do not find any systematic behavior, while for very hot systems (T>5 keV) the spectroscopic measurement may be strongly overestimated up to a factor of 4; iv) Si is well recovered for all the clusters in our sample. We investigate in detail the nature of the systematic effects and biases found. We conclude that they are mainly connected with the multi-temperature nature of the projected observed spectra and to the intrinsic limitation of the XMM-Newton EPIC spectral resolution that does not always allow to disentangle among the emission lines produced by different elements. Comment: (e.g.: 17 pages, 8 figures, accepted for publication in the Astrophysical Journal, updated discussion to match published version-new section:6.3)