Thanks for all the help guys.
The answer is NO.
The POE in via the ether1 interface is NOT directly connected to the J9 power input jack. They each have (apparently) separate reverse polarity protection diodes. So the power coming in via the POE does NOT appear on the J9 jack. Likewise the reverse is also true; the power in the J9 jack does NOT appear on the POE connector.
While I can understand the reverse polarity protection, I do not fully agree with separating the inputs.
They could have done the whole thing one better by replacing the reverse-polarity protection diodes with a bridge rectifier. Then the input voltage could be automatically correct reverse-voltage input.
They've been doing this on POTS phones for years...
There is a pair of pads on the circuit board right behind the POE/J9 connectors labeled D8 that has the input voltage on it right after the diodes. That makes a fairly decent place to tap the voltage. I took a phone-snap of the final result:
The idea here is that we bring the WAN in from a Canopy SM via the WAN PWR connector in the surface-mount box to the left of the RB133. That RJ45 splits the power and data. The data goes to the ether2 interface (labeled WAN), and the power comes from the two pads just above and between J9 and the POE connectors (it sneaks in between the serial port and the POE jack). The wlan goes to the repeated site(s).
The slick part about this setup is that there are NO external boxes holding anything. The POE connector goes to the local subscriber, and gets power for both the MikroTik _AND_ the Canopy SM. A short pigtail goes to the Canopy SM, which is mounted a couple feet away from the MikroTik. That's the whole gig; two radios outside the premises, one of which is both a router and an AP.