Bad redshift determination with version 0.22.0?

The simple sanity check on the redshift determination in the recent end-to-end processing (see attached) seems to show strangely bad redshift determination. The redshifts from 1D pipeline are mostly ~0.9 or ~2.3. Could you confirm that this is real or I'm doing something wrong? I'm using version 0.22.0 and the configuration is attached. The results are also in https://pfs.ipmu.jp/internal/devarch/ipmu/pipe_e2e/ .

check_redshift_determination.pdf

^{parameters_e2e.json}