Thanks for the reply! But no, that's not it. I tried altering Tinit from 1.9 to 2.5 msec in 0.1 msec increments and the results are always the same:
In fact, nothing changes if I leave out the wake-up pulse:
... which makes me wonder if the chip in fact wakes up. But even if I send it a huge honkin' 500 µsec wakeup pulse, nothing changes:
Obviously I must be doing something wrong, but what? The only other rational explanation is that the chip is possessed.