Tom, the problem is that you don't know the torque. You know the starting torque because it is the line-start torque reduced by the square of the starting voltage reduction. But, as soon as the motor starts to turn, two things happen to make the torque difficult to predict.
First, you have the usual NEMA B torque speed curve which, at any constant voltage including reduced voltage, is far from linear, much less horizontally flat!
Second, the softstart is ramping up the voltage as the speed increases so, on top of the above, you have the voltage steadily moving back to full line voltage. That puts a decidedly upward slope on the already non-linear torque curve and, if that isn't bad enough, you don't know just where in the torque-speed curve the voltage reaches full line level since the voltage ramp is based on time. I've seen the time set so short that full voltage occurs before the speed reaches break-down torque level and I've seen time set so long that full voltage isn't reached until several seconds after the motor hooks up in near-synch mode and is running near normal speed (lightly loaded).
All of that makes accel time on a softstarter nearly impossible to calculate and high risk to predict! Field trial is the only reliable way I've found to determine accel time and, even then, you are never sure the test load torque matches true running conditions.
Softstarters are nice but precise they are not! For precise you need an inverter with an enforced accel ramp.
