APC UPS magnetizing current and standby power. SMT1500

[bdunham7]

I'm renovating some UPS APCs, including this SMT1500 1.5kVA/1kW sine-wave model.  It was working except the batteries were bad.  I reprogrammed it for a slightly lower float voltage and then I wanted to see how that affected standby power consumption since a similar model that was in an enclosed cabinet was getting warm and had expanded batteries.  So when it was all done, I measured the power and current after it was fully settled in and I'm not sure I can explain why it is doing what it is doing.

The SMT1500, like most of these block-shaped sine-wave SmartUPS models, uses a large-ish EI transformer with a nominal 24VAC secondary and a 120V input with two extra taps at ~110 and 130V.  It uses these for buck/boost during slightly low or high voltage--they call this "AVR trim".  It also uses a MOSFET H-bridge to drive the transformer during backup operation and the same MOSFETs as an active bridge rectifier for normal operation. 

 I set up a current clamp with a scope, a Kill-A-Watt monitor on the input and a DMM on the output.  No load was attached.  With the batteries fully charged, turning on the UPS and letting it settle results in a power reading of ~14W, a current reading of 0.65A and a VA reading of ~77VA.  V_IN = V_OUT = ~120V +/- 2V.  Battery charge current is pretty low, <5mA.  This seems pretty much as expected, with 0.65A a reasonable magnetizing current for a pretty big transformer, a few watts lost to copper and iron losses in the transformer and a few more to run the UPS controls and fan.  I was considering putting a capacitor across the transformer to reduce the overall reactive current.  This probably isn't necessary since 0.65A isn't much recirculating current, but it wouldn't hurt to reduce it, right?

However, after sitting there doing nothing for an hour or so, there is a 'click' sound and I see the power reading drop to ~8W (good) but the current goes to 0.95A.  The input and output voltages are still nominal and identical.  The only reasonable explanation for this AFAIK is that the UPS is switching both the input and output to the 'shorter' tap on the transformer.  The question is...why?  What does this accomplish?  And does the input power drop because it does this or are there just two things going on at once?

Any ideas?  Scope shots of input current are attached, both before and after the transition.  I included AVG(64) and ERES(3.0) traces to show that the higher-frequency components are not just noise.