4160V Chiller Motor Start-Up: Auto-Transformer vs Primary Reactor, Voltage Dip Causes

A 4160v chiller motor is drawing too much power at start-up. Utility company only lets them run it once or twice a year because of it. It dims all the lights etc.. It was originally a primary reactor starter (2 contactors). The PR was replaced with an auto-xfmr and wired the same. So now it’s a series auto-xfmr with no shorting contactor. The xfmr is fed, the secondary output supplies the motor at start-up, then a bypass or run contactor pulls in. So the xfmr is not shorted or disconnected from the circuit. 
What could be causing the voltage dip on the source, or why is it causing it?
And why does a 3-phase auto xfmr need a shorting contactor? What does connecting the windings do besides dissipate the energy in the winding, and what does leaving them open do?
Thanks!

Hi Kenneth-A series reactor or autotransformer will allow the motor to "soft start" to reduce the inrush when the motor starts. The series reactor is much more common than the autotransformer.  I have seen installations where the series reactor was not shorted after the motor was running normally, but if an autotransformer is not removed from the circuit after the motor has started the motor will be running at a reduced voltage continuously.  Why was the reactor replaced with an autotransformer?  It's possible that the output voltage of the autotransformer is set too low, and the motor is never coming up to full speed.  Do you have a tachometer? If you measure the speed of the motor and compare it with the nameplate speed that could give you a clue. Also- when you said the autotransformer is wired the same as the reactor- the reactor would have two wires per phase and the autotransformer would have three.  I wonder if the autotransformer is being used as a series reactor with only two wires connected? -Rick