Because he put so much weight on trusting our senses, it's hard for me to imagine he would propose belief in a prolepsis which violated sense data.
I think this is an essential part of the picture we can't lose sight of. The three legs of the canon are supposed to function together, and while the data received from all is entitled to respect, the opinions/conclusions we accept as "true" have to be tested and conformed to all three. Now sometimes we aren't going to have as much, or as credible, data from some as we have from the others, and so we're constantly concerned about when to "wait" and when to consider an opinion as sufficiently confirmed to act on it or consider it to be confident.
But the reason I single out that quote is that I definitely agree that where we have clear evidence from one leg of the canon that actually contradicts the evidence from another leg, we definitely should not consider such a state to be worthy of considering it to be confirmed. There will be perhaps lots of instances where we have some evidence from some legs that isn't strong enough to be confident of on its own, combined with *no* evidence from the other legs. Then we have to treat the situation in its own context and decide whether we have enough at that point or need to "wait." But where we have some affirmatively contradictory evidence from one of the legs, I presume that at least in most cases that means that we definitely don't have enough evidence to consider the matter settled.
All of which leads to an ad-hoc reasoning process that doesn't allow for magical bright lines, but I think that's just the kind of "reality" that an atomistic universe implies, so it's probably the right position.
EDIT: I suppose we have to go even further and say that there might be *some* evidence that contradicts the rest, like in some kind of sensation trick where we are blindfolded and have to speculate without sight (I am sure some of you can easily come up with an example). But even in those cases, the contradictory evidence is generally temporary, and can be ultimately corrected when we take the blindfold off, or get more data. So I would think in most all cases we can still say that we wouldn't accept a conclusion where there is contradictory evidence that isn't limited in time or scope or in some other way we have evidence of its limitation, by which we can explain the contradiction.