Multi-reservoir systems based on theMacroscopic Fundamental Diagram (MFD) look appealing to simulate traffic states at large urban scales. Nevertheless, flow limitation due to spillbacks between reservoirs is still not fully understood yet. Recently, we analyzed flow exchange properties and formulated new merging and diverging models, different from the ones previously proposed in the literature. However, both the latter and our approach received very little support from aggregated link-scale data (real or simulated one). The contribution of this study is to validate different MFD-based modeling approaches by using microscopic simulation. To this end, we focus on an artificial Manhattan network crossed by two regional paths (West-East and North-South). We then analyze the aggregated accumulation, inflow and outflow of these paths for different congestion scenarios, and compare the results with the outputs predicted by the MFD-based models. It is notably found that the outflow diverging scheme is critical to fairly reproduce microsimulation results. During congestion on-set, the widely used approach of a decreasing outflow demand with independent partial outflow treatment provides good estimation of transient states, but fails to predict reliable network unloading during congestion off-set. The modified approach we developed in our previous study takes advantage of setting the reservoir outflow demand to maximum and applying inter-dependency relationships between outflows to overcome this issue. To a lesser extend, using a demand pro-rata merging scheme, as it is usually done in the literature, would be the best option to determine partial inflows properly.