Bad reconstruction for short scan of near parallel geometry

[SimonRit]

The following does not produce an accurate result:

rtksimulatedgeometry --sdd 1701 --sid 1625 -o g.xml -n 100 --arc 180
rtkprojectshepploganphantom --spacing 0.23 --size 200 -g g.xml -o p.mha --phantomscale 20
rtkfdk -p . -r ^p.mha$ -o fdk.mha -g g.xml --spacing .2

as if the first and last projections had a lot more weights than anticipated. Doing a purely parallel geometry is also not working

rtksimulatedgeometry --sdd 0 -o p.xml -n 100 --arc 180
rtkfdk -p . -r ^p.mha$ -o fdk.mha -g p.xml --spacing .2

as the resulting image is filled with NaNs.

[SimonRit]

The parallel geometry works with the --nodisplaced option in rtkfdk. Not sure how this could / should be automatically triggered. The divergent geometry is due to Parker not working at all with short scans smaller than $\pi$. Not sure either how to avoid this.