If so, I do not see how.
Suppose P infallibly knows A. The definition specifies that infallibly knowing is a statement about all possible worlds, namely that P believes what ever is true about A in all possible worlds. Therefore, if we assume that P believes what ever is true in all possible worlds, then they believe that A is false in one possible world W, wherein A is false.
It is only a contradiction, if we say that “infallibly knowing that A holds” implies “knowing A holds”, but by Def. 2 it does not. That was my entire point. That if we define infallibly knowing in a way that still permits for the infallibly known statement itself to be false in some possible worlds, then that costs us the ability to infer knowing from infallible knowing, which… well, for lack of a better argument, feels weird.
One could, of course, construct one (arguably somewhat artificial) remedy for this, in this way:
Definition 6: Entity P is said to ‘infallibly know’ proposition A only if A is true, and in all possible worlds W: P in W has knowledge of the truth-value of A in W.
This is, basically, Definition 2 + a stipulation that A is true in the world whence the statement is made. This satisfies the possibility for A to be false, maintains P’s knowledge about A in all possible worlds. The price paid for this, is that now the statement “P infallibly knows A” can only be true in worlds where A holds. In a world where it does not, they may know that it does not hold, but they cannot have infallible knowledge of A’s truth-value, because we have defined infallible knowledge as including the truth of the proposition in question. They can, however, know that not-A in such worlds, interestingly enough.
Now that I think about it, perhaps this is a more sensible interpretation than I figured so far… 