We use the PyNeb 1.1.16 Python package to evaluate the atomic datasets available for the spectral modeling of [Fe II] and [Fe III], which list level energies, A-values, and effective collision strengths. Most datasets are reconstructed from the sources, and new ones are incorporated to be compared with observed and measured benchmarks. For [Fe III], we arrive at conclusive results that allow us to select the default datasets, while for [Fe II], the conspicuous temperature dependency on the collisional data becomes a deterrent. This dependency is mainly due to the singularly low critical density of the 3d7 a4F9/2 metastable level that strongly depends on both the radiative and collisional data, although the level populating by fluorescence pumping from the stellar continuum cannot be ruled out. A new version of PyNeb (1.1.17) is released containing the evaluated datasets.