RICH sector 4 pixel positions change when sector 1 RICH included

[cpecar]

The ccdb has a bank /geometry/RICH/setup which determines which sectors contain a RICH/where RICH 1 and 2 are located. However, it seems that adding a second RICH sector somehow alters the position of the first RICH.
Using a copy of the ccdb with RICH alignments set to 0 and the only change being to /geometry/RICH/setup, the results of a few pixel centers dumped using RICHGeoFactory are:

Pixel positions, sector4 RICH, no RICH in sector 1:
pmt 7 pixel 1 Point3D: -50.28602 -13.76300 690.86778
pmt 7 pixel 2 Point3D: -50.28602 -14.38000 690.86778
pmt 15 pixel 1 Point3D: -55.08945 -11.11300 688.62790
pmt 15 pixel 2 Point3D: -55.08945 -11.73000 688.62790

Pixel positions, sector4 RICH, RICH added to sector 1:
pmt 7 pixel 1 Point3D: -50.85337 -13.13700 690.86778
pmt 7 pixel 2 Point3D: -50.85337 -12.52000 690.86778
pmt 15 pixel 1 Point3D: -55.65680 -10.48700 688.62790
pmt 15 pixel 2 Point3D: -55.65680 -9.87000 688.62790

Marco and I came across this when trying to cross-check pixel positions between the gemc and coatjava implementations of the RICH geometry.