Numerous studies in packed-bed biomass combustion modeling have been carried out, with the validation being usually performed against experimental data obtained in regions far from the bed. It is known, however, that in these regions standard modeling has a tendency to overpredict the solid and gas temperatures and the mixing rates of the gaseous species released from the packed-bed. This leads to an underprediction of the unburned hydrocarbon and soot concentrations in these regions, originating incorrect estimates of the pollutant emissions. Due to more and more stringent legislation, the combustion appliances need to improve their designs, for which deeper studies of the near-bed region modeling are necessary. This study seeks to find the main causes for the temperature and gaseous species misprediction in this region. Current biomass combustion models represent the gas solid interaction by means of a porous medium, impeding the representation of the streaks with nonperfect mixed species that are known to be present. The modeling of gas-phase reactions with the standard eddy dissipation model is not suitable to account for complex chemical mechanisms and the weakened turbulent region created above the bed, despite its computational robustness and cost being highly appreciated. This work presents the first step toward the formulation of a modified Magnussen parameter (A), so the reaction rate resulting will reach a better physical modeling under transient packed-bed conditions, recovering the original value proposed for the Magnussen constant. In this work, the best adjustment was found for A* = 2, instead of A = 0.6-1 as proposed previously by other authors, especially in regions far from the packed-bed. As a result of this work, the computational fluid dynamics (CFD) code was able to predict the spatially resolved profiles of temperature and major gas species concentration in the flame region of a domestic boiler with reasonable accuracy.