There is mention in Lucretius about how images floating throug the air strike us, even when we are asleep, and that these images striking our brains (at least, not through our *open* eyes) create stimulation that leads to dreams and sometimes thinking we see things that we really don't.
(1) One problem is if people use this to conclude that we are seeing pictures of real gods moving among us. That would be wrong and would be a "false anticipation" because we know from other reasoning about the gods that they are perfectly blissful, and in the intermundia, and so there is no way we are going to see them moving around us. Even if we received images all the way from the intermundia (which apparently there was speculation that we do), those images would only be of blissful and perfect beings who have no interest in us.
(2) Another problem is that some later Epicureans elevated this process of receiving images, forming a concept, and then comparing those concepts to what we see in the future. (I think DeWitt speculates probably in response to Stoic drumbeating about the importance of logic.) And they considered this extended process to be a "fourth leg of the canon" in addition to the three that Epicurus had stated. The error here is that "concepts" are formed in our mind only after we make judgments about what we get from the three canonical faculties, and the process of making judgments involves opinion and the possibility of error. Thus although this process of forming concepts is very important and useful, it is not safe to call it a "test of truth" because the concept we are using could be erroneous. Yes any sensation can be erroneous, but we trust the canonical faculties to be programmed by nature and not by ourselves. Concepts formed by our judgment/opinion could be totally erroneous and are not subject to the same natural correction mechanism as for example "looking twice" would do it for the eyes.
The second issue is a generalization of the first. We would conclude that a concept of an actively intervening god is a "false opinion/concept" because it is not based on any proof and it is contradicted by other evidence (our other observations from images and our reasoning about things that are "perfect"). If the person arguing for the false concept of an active god is alleging that he saw a vision of a god visiting him, then we can't necessarily rule out that he saw "something," because all sorts of images are flying through the air. But we would conclude that his perception of the image was distorted or a "false anticipation" in the same way that we see images of oars in water being distorted. Just like with vision or any other sense, a perception from a canonical faculty is reported "truly" (honestly) but that is not to say that the perception conveys all the facts accurately.
Those are the two issues I see being raised most often about the flow of atoms and how we process them. Going in the direction of a "fourth leg of the canon" has the potential to let the nose of dialectical logic into the canon. The people who think that the existence of images can be used to prove the existence of active gods are not considering that these images can be distorted, and distorted images can be just as false to the facts as a view of a tower that appears round at a distance but is really square.
Note: Letter to Menoeceus (Bailey) For gods there are, since the knowledge of them is by clear vision. But they are not such as the many believe them to be: for indeed they do not consistently represent them as they believe them to be. And the impious man is not he who popularly denies the gods of the many, but he who attaches to the gods the beliefs of the many. For the statements of the many about the gods are not conceptions derived from sensation, but false suppositions, according to which the greatest misfortunes befall the wicked and the greatest blessings (the good) by the gift of the gods. For men being accustomed always to their own virtues welcome those like themselves, but regard all that is not of their nature as alien.