But I see I was making a mistake in being so specific. But I think Don has a good point about them being "always true" in their role as primary ways of knowing.
Respectfully to all concerned let's state all this in a way that is more clear about the "always" and the "absolute" implications. I think the point Epicurus was making is that "YES - "anticipations" as a PRE-conceptual source of raw data are always reported "honestly," but "No! - anticipations are not themselves ethical conceptions / propositions and they are themselves never fully formed conceptions that are absolutely applicable in all circumstances!"
Due to the nature of the universe (no gods, no ideal forms, nothing that is "always the same" except the fundamental elements themselves) there is no way possible that any ethical concept can ever be absolute for all people, all times, all places. That is in my view the clear meaning of the "justice" statements in the last ten PD's, but it's also compelled by the Epicurean physics. That's where DeWitt slipped (in my humble view) in describing anticipations as "ideas." They are used to FORM ideas, just like the data from all three of the canonical legs are used in our minds to form ideas, but they are not ideas themselves.
This is the point that was the fatal flaw in those who adopted a fourth leg of the canon (as cited in Laertius) and it's the flaw committed by Cicero's Torquatus in saying that he was one of those who believed that the proposition that pleasure is desirable should be the subject of essential logical proof.
What we're discussing here is what (in my opinion, I think following DeWitt) blew up the Epicurean movement in the ancient world. They gave in to the Platonists / Stoics and accepted the argument that their philosophy required proof through "logic" - when Epicurus told them that "logic" is not something that is canonical, and that proof comes through the natural faculties (data from all three canonical faculties viewed generally as "sensations") rather than through mental conceptions / logic.
I know pleasure because I can feel it, not because I can define it absolutely accurately in words. The same would go for all the inputs from all the canonical faculties. The data does not come to us in words/concepts, those are just devices that we use to try to describe them, but the words are just devices. The words cannot be mapped one-to-one exactly to the full context of the experience.