To date, all T counters are timing to within 1 ns of each other. The spurious T channel 17 is well timed, but a significant trigger cable discrepancy length pushes non-overlap events in T counter 9 farther in time. However, this doesn't affect overlap Tchannels 16 and 18, since the `faster' T counters 8 and 10 determine the timing of those overlap regions.
A few days ago, I created the calibration database E_correction _factor, and filled it with the g10 values. I also checked code into CVS that reads in and applies these correction factor to the original map.
A more rigorous study (more rigorous than eyeballin', at least) has been done suggesting that the new values above are not perfect, but very good.
This study, which will probably become a PrimEx Note, is located here.
This note is essentially complete, and I submit it to the collaboration for consideration.