In cold astrophysical environments, some molecules are observed in the gas phase whereas they should have been depleted, frozen on dust grains. In order to solve this problem, astrochemists have proposed that a fraction of molecules synthesized on the surface of dust grains could desorb just after their formation. Recently the chemical desorption process has been demonstrated experimentally, but the key parameters at play have not yet been fully understood. In this article we propose a new procedure to analyze the ratio of di-oxygen and ozone synthesized after O atoms adsorption on oxidized graphite. We demonstrate that the chemical desorption efficiency of the two reaction paths (O+O and O+O$_2$) is different by one order of magnitude. We show the importance of the surface coverage: for the O+O reaction, the chemical desorption efficiency is close to 80 $\%$ at zero coverage and tends to zero at one monolayer coverage. The coverage dependence of O+O chemical desorption is proved by varying the amount of pre-adsorbed N$_2$ on the substrate from 0 to 1.5 ML. Finally, we discuss the relevance of the different physical parameters that could play a role in the chemical desorption process: binding energy, enthalpy of formation, and energy transfer from the new molecule to the surface or to other adsorbates.