"The dc current becomes much less distorted. I had hoped that this would benefit alternator efficiency as they really don't like rectifier loads. In fact I saw no improvement in overall efficiency and didn't chase this mode."

I would also think that an improvement in efficiency is possible using this scheme. I wonder why you, in fact, didn't see it.

If you're able to derive current from the portions of the input waveform that are below the main rectifier cut-in, it would seem that total output should increase.

Of course, the closed-loop bandwidth would have to accomodate the alternator frequency  to allow the PWM to follow the input waveform. Given the high ratio of PWM frequency to input frequency, this would not seem difficult.