Depending on which Voyager xyz trajectory file you're using, it might not have enough samples around the times of the various flybys. Unfortunately, Celestia only does a linear interpolation between adjacent xyz samples. If the samples are too far apart, the spacecraft will seem to take a shortcut instead of flying around a planet.
For a while even Chris thought there was something wrong with Celestia when Galileo was missing the Earth by 100,000 km during its gravity-assist flyby. Galileo seemed to be bypassing the Earth on the opposite side from where it actually went. This was simply because the samples weren't frequent enough for that part of its trajectory. Once the frequency of samples was increased from 1 every 2 days to 1 every 5 minutes or less, Celestia showed Galileo following the right path.
For what it's worth, the part of Voyager 2's trajectory past Neptune is available on my Web site with 15 minute sampling intervals. Using this, Celestia v1.2.5 does an excellent job of reproducing the Triton flyby.
See
http://www.lns.cornell.edu/~seb/celestia/spacecraft.html#3.4.10
