The perils of stacking optically selected groups in eROSITA data. The Magneticum perspective
Hydrodynamical simulation predictions are often compared with observational data without fully accounting for systematics and biases specific to observational techniques. Using the magnetohydrodynamical simulation Magneticum, we generate mock eROSITA eRASS:4 data, combined with GAMA-like spectroscopic surveys and optically selected galaxy catalogs from the same light-cone, to analyze hot gas properties in galaxy groups via a stacking technique. This study aims to (i) incorporate observational systematics into predictions and (ii) evaluate the reliability of stacking techniques for determining average X-ray properties of galaxy groups. Our analysis provides X-ray emission predictions from Magneticum, including contributions from AGN, X-ray binaries (XRBs), and the Intra-Group Medium (IGM) as a function of halo mass, covering Milky Way (MW)-like groups to poor clusters. We find that AGN and XRBs dominate the X-ray surface brightness profiles of low-mass halos. The reliability of stacking techniques is tested by reproducing input X-ray surface brightness and electron density profiles, accounting for completeness and contamination of prior samples, miscentering of optical group centers, uncertainties in X-ray emissivity due to gas temperature and metallicity assumption, and systematics in halo mass proxies. The halo mass proxy emerges as the primary source of systematics, affecting X-ray surface brightness and scaling relations. We show that stacked X-ray luminosity-mass relations are flatter than input relations but consistent with observations. Additionally, the retrieved hot gas fraction-mass relation aligns well with observational data. These results highlight the need to account for systematic errors when comparing stacking techniques to other methods using different prior catalogs or predictions.