Singular matrix in spectral extraction

rhl reports:

With

reduceExposure.py /work/drp --calib /work/drp/CALIB-PFI-20211220 --output /work/rhl/rerun/rhl/may-2022  -c isr.windowed=True isr.doFlat=False doAdjustDetectorMap=False doMeasureLines=False --id visit=76003..76007 arm=r

I get this , but the 76003 data looks OK.
the point being the column=1025/1051 dropouts. It looks as if the only interesting bit is NO_DATA

On inspection, the matrix for column=1025 is singular, due to the same single pixel being available for fiber indices 515 and 516.