My belief isnt mistaken, but yes, it looks like you are correct that it doesnt happen through a diode bridge from the smoothing cap, on my own re-inspection. I hadn't looked at it that hard before, but it looks like one of the diodes conducts, when the other doesnt, and that one that isnt conducting stops the cap from self-discharging behind.
But it can happen in different machines than this, there's always the possibility that a bypass cap is going to charge back down the machine as well as in front.
It comes up when you have a "multi-rail" when your powering more than one unit off the same battery, it can happen there too, and can just go straight backwards through the diode, given the specific situation.