Notes. The simulations here all assume a catastrophic failure fraction of 35%. Combining this rate with the observed outlier rate in COSMOS and the spectroscopic number counts in KiDS gives the expected failure rate: 1.30%. The simulation is slightly shallower than the KiDS data, giving an outlier rate of 1.03%. We simulate our outliers with this rate (Fiducial) and with two artificially inflated rates: increasing the instance of failures over all magnitudes, and increasing failures at faint magnitudes. The three rows in each set of rates show the results when forcing failures to be brighter than the target galaxy (z ∼ N(z|i ≤ i₀); top), arbitrarily sampled from the full spec-z distribution (z ∼ N(z); middle), or uniformly sampled from the redshift range (z ∼ U(z); bottom). In all cases the results show shifts of less than 0.005 in the tomographic bin means. Given the magnitude of these biases is small, we conclude that catastrophic spectral mis-identifications cannot cause significant biasing of the SOM tomographic redshift distributions for KiDS-like analyses.