We might start from some concrete examples of his assertion. I know where my keys are, and I am wrong but not lying. I am confused. Or I know dinosaurs aren't birds, and I am wrong, but not lying. I was taught biology 30 years ago, and I am misled.
To some degree, in each case, I have done the work of justification and succeeded at establishing the truth of the proposition, but then the world changed. So the issue is not truth, or justification, it is time and mutability.
Beyond the uncertainty the quote points directly at, having these additional dimensions makes it difficult to classify knowledge as a mental state.
You could go well out of your way and identify it as a component of a composite mental state that transforms across time. But then what isn't? The notion is dauntingly broad. And you would need a separate component for every nuance of every item of information that you separately know, which takes us in unhelpful directions.
At that point, the question is answered, without needing a real definition of knowledge in terms of mental states. But looking deeper might help us consider the reason for the confusion.
(The quote is clearly dealing with knowledge as a personal property, and not on knowledge as a transmissible currency. It is to that latter form that the notion of 'justified true belief' applies better. And it is the former form that I will seek to define.)
We know (vaguely) when we come to know a proposition, and subjectively that point is a moment of belief. So there is motivation to consider knowledge to be the temporal trace of a past belief.
But there are confusing difficulties with knowledge of propositions that other mental states also generally don't have, which give it a mechanical feel. I can both know X's phone number and not be able to retrieve it at the moment. To the degree that I am just re-entering a past mental state, this should not be possible.
So the trace itself is not an image, it is more complex.
I can also only vaguely know something I once more clearly knew. If knowing were just stored belief, this fuzziness of knowing should be the same level of indecision that the belief had when I initially held it, or should at least be related. And it does not seem to be. The fuzziness of time seems independent of the surety at learning.
Beyond that, there are things we know that are not propositions. We know procedural things: driving a car, playing the piano, speaking... And we know impressionistic things: how to choose the right gift for the right person, when to use what rhetorical tone...
A lot of that is clearly knowledge. But because of its form, we are seldom conscious of it. And we may not consider the decision we made to act in the way that constitutes the knowledge itself to be a belief in particular, but it does involve some assent of faith.
Without some internal model of what memory and processing are and how memory modifies itself, the best you can honestly propose as a definition is that knowledge is the effect that faith in a past decision has upon your future thinking.
The most clearly identified effect would be reuse of information captured in a past mental state by reflecting upon or partially reconstructing the times you have decided that information had value. And that is probably why we think of knowledge as a mental state itself. But it is hardly the most important one, much less the only one.
Adopting a sort of Jungian psychoanalytic model of memory as a network of cathectic complexes, knowledge can be characterized as the latent content of an inactive potential mental state.
A memory forms connections to the memories around it, which reshapes them, but most of the time it just lies there until a related piece of knowledge makes it relevant, at which time it reactivates as an image of the memory of acquisition and information merged with that memory at each prior activation. (Complex formation is a form of Hebbian learning that takes place during cathexis.) Given stronger activation, the image is realized and one can temporarily re-enter the modified prior state.
