We reconsider the long-standing question of the critical defect hopping rate rc in the one-dimensional totally asymmetric exclusion process (TASEP) with a slow bond (defect). For r < rc a phase separated state is observed due to queuing at the defect site whereas for r ≥ rc the defect site has only local effects on the stationary state of the homogeneous system. Mean-field theory predicts rc = 1 (when hopping rates outside the defect bond are equal to 1) but numerical investigations seem to indicate rc ≈ 0.80(2). Here we improve the numerics to show that rc > 0.99 and give strong evidence that indeed rc = 1 as predicted by mean-field theory, and anticipated by recent theoretical findings.