Redshift and wavelength offset between the observed [O III]λ5008 emission lines measured from the PRISM and G395M spectra. Taking the G395M spectrum as ‘truth’, we found a systematic offset of Δz∼0.0044 or ∼0.25 detector pixel for the PRISM spectrum that does not appear to depend significantly on wavelength (grey solid lines show the running median). The scatter can be partially explained by the larger uncertainty for fainter emission lines. In addition, the intrashutter position of the source (i.e. the spatial offset in the dispersion direction) also introduces wavelength offsets of up to 1 pixel, if the source is point-like and located at the edge of the shutter. In practice, high-redshift sources are (moderately) spatially extended, resulting in smaller offsets. We indeed find a correlation between the source position in the slit and the wavelength offset.