Epiphenominalists are often trying to accommodate scientific facts without discarding the reality of the mind. Various things happen that do not accord with the folk-psychology that all immediate experience must be shaped the way we generally discuss it in words. Instead, it is partly conditioned by the existence of words, and their purpose.
(The entire motivation comes from looking at the proper boundaries between different sciences. It makes as much sense to discuss the mind in terms of brain circuitry as it would to discuss baking in terms of the shapes of orbitals making up chemical bonds. It is two layers of epiphenomenon away from what is reasonably tractable -- it skips over both behavioral conditioning caused by and through language and social construction by acculturation. So arguments that make that leap and put the two side-by-side end up being misleading or vacuuous. This includes philosophical inquiries about what things like qualia "are, physically".)
From an evolutionary point of view, feelings are about their survival consequences, not about their causes. In particular, we see in psychotherapeutic contexts that feelings and impressions do not arise immediately and unconditionally at the point where the events that cause them, happen. That would mean that the brain feels what it has evolved to feel. It does not actually feel its own state. It feels the emotional or informational tendencies that those states predict should be most efficacious given its evolution. The feelings are epiphenomena and not an expression of a given state of current reality within the brain or within the environment at all. It is not unreasonable to generalize that observation to all "felt states".
The behavior of people raised in deep isolation, and the memories of people newly given language and social exposure later in life indicate that beforehand, they do not experience ideas in the same way. For instance, navigation through space, something we think of deeply intuitive and 'built-in', improves when people are first given words for directions and kinds of movement long after infancy. This suggests that ideas, even ones as basic as our sense of geometry, which we often use without consciously thinking in words, are deeply related to the ability to express them to other people.
Even things we do not think of as stories may take the form of the preparation of social narrative, a draft of a description of our state, (if not in language, then in some related social product we could deliver or enact if asked,) as in Dennett's "multiple drafts model". Taking this to its logical extreme, if increasing expressive options makes this much difference, without other people and their social expectations, encountered early in life, it becomes doubtful we would experience the "having of ideas" at all. So the whole notion of thought that is built on describing it in terms of ideas cannot be taken for granted as natural. But we automatically do that when we assume the reality or nature of qualia and consciousness -- because those terms are conditioned on self-observation in terms of ideas. If those are created by the social contract, we do not need a naturalistic "brain-level" description of them. And giving one would be inappropriate.
The insights of modern perceptual experiments on reaction timing suggest that qualia and the experience of time passing take time to appear in the mind. So they are most likely not fundamental, but derived from our reasoning process. This means the brain does not know about qualia, it knows about the labelling of experiences. It does not know about consciousness, it knows about extremely-short-term memory. We invent these more abstract terms, but what we can really identify of them always reduces to something less esoteric. What we can identify of consciousness is the stream of language and experience that marks the passage of time. What we can identify of qualia are memories of images or sensory traces and qualities we associate with them. Once you step away from the forced concept into the behavioral contents, the mystery disappears.
All of this paints a picture of 'immediate experience' that is much less immediate and is instead a learned game. You can have certainty that these things exist in the same way you can enumerate the qualities of a unicorn. You can know the socially acceptable behavioral correlates. The actual experience is just a finely tuned method of evoking those correlates on cue.
So the answer is that we cannot explain this knowledge because there is no such knowledge. There is training to conceptualize things in this manner to improve our social interactions. They are not information, they are learned habits of reference, that does not mean they actually refer to anything.
