Rick,
I changed PVset to 0.8V by putting as parallel resistor of 3k9 across the 1k on the PVset divider chain.
The sawtooth reduces to 3.6V p-p, and the dc voltage (at no drive) on both the CSout and Aref pins drops from 2.83 to 2.53 V. So far so good.
I can now get 6V output drive in one direction by changing the drive into the INV/COMP pins. But nothing in the other direction.
However, when I started off, I tried just holding the 3k9 resistor on by hand and found that when I removed it, the whole chip started working just fine and I could get 11V drive in both directions!
Cycling the power returned it to the initial unipolar state, but the process could be repeated.
It seems that selecting a 5V p-p sawtooth with a 2.5V reference causes some sort of internal latch-up.
We have now decided to stick to using the 6V centre tapped Arefin and that works just fine.