A benchmark neutron propagation experiment in iron was used to compare experimental and calculated reaction rates to test ENDFIB iron cross sections. Excellent agreement was found between experiment and calculation when ENDF/B-I data and a more recent Oak Ridge National Laboratory evaluation were used. A background effect of the manganese impurity, stronger than earlier expected, is shown to play an important role in the assessment of the 25-keV s-wave scattering resonance minimum. The deficiencies in the high-energy (>30 keV) range of ENDFIB-III data, which are also indicated, seem to be overcome by the most recent evaluations.