Phenomenal Properties and Epistemic Access

Phenomenal properties are properties of mental states. In virtue of a phenomenal property a certain mental state feels somehow, there is something it is like to be in that mental state.
Some philosophers have argued that a mental state M of a subject S can instantiate a phenomenal property P without S realizing (or even being able to realize) that she is feeling anything (phenomenal consciousness without access consciousness in Block's terminology).
I disagree. There is a sense of feeling, that is the sense I am interested in, in which it makes no sense to talk about feeling anything if one does not realize it. In that sense, phenomenal consciousness entails access consciousness.
If we are interested in phenomenal properties and in its naturalization, the discussion is relevant. For imagine that one is interested in a neural correlate of a conscious mental state, or in some empirical evidences relevant for certain theories of consciousness. Is the epistemic access a constitutive part of the phenomenal property?
For instance, blindsighters have been sometimes presented as an objection to representational theories of consciousness. In order to deal with this, representational theorists introduce some further condition for instantiating a phenomenal property besides the representational character (for instance Tye introduces the condition of being available for reasoning and believes -being "poised" in Tye's terminology). But if we accept the distinction between the phenomenal property and the epistemic access, we can say that what is missing in the case of the blidsighter is the epistemic access (poised would not be a necessary condition for consciousness). In that case, I see no pre-theoretical way to decide whether or not a phenomenal property is instantiated.
A further problem would be that, if the process responsible for the instantiation of the phenomenal property and the epistemic access are different, one could fail. Imagine that S is instantiating phenomenal property A, the epistemic access machinery fails (certain neurons misfire) and indicates phenomenal property B. What does S feel? Trilemma:

1. S feels anything. But this seems to be an ad hoc answer
2. S feels B. In this case the phenomenal property instantiated plays no role in what S is feeling.
3. S feels A. In this case S feels A but if she has a believe about what she is feeling this is going to be false. This option seems to go against the widespread intuition that we do have direct access and knowledge of what we are feeling
   

Option 3 seems not to be acceptable for me. One can fail in categorizing the feeling: for example having a experience A of very cold water and believing that it is really hot. In such a case there is a categorization mistake: experience A is categorized as belonging to experiences of hot water. Nevertheless, it seems to me that I am infallible in knowing what it is like to have experience A when I am undergoing experience A (some kind of indexical knowledge, this feeling)
It seems to me that in virtue of instantiating a phenomenal property I thereby come to know what it is like to undergo the corresponding experience (maybe I cannot remember it 1 msec. later). If this is true, the epistemic access is an intrinsic element of the phenomenal property and there cannot be phenomenal consciousness without access consciouness.

What do you think about the relation between phenomenal properties and the epistemic access?
Thanks