The answer of Locke is given in the abstract of the paper. We are practically certain if and only if
the actual epistemic probability that p differs from epistemic certainty that p only in ways that are irrelevant to the decision one currently face.
I.e. if our decision is based on knowledge/expectations/perceived certainty we may well be wrong in our assumptions and it is asked too much to base decisions on an "absolute" certainty that cannot be achieved anyways. As long as the points and extent of that wrongness would not make a difference for the outcome of our decision process, we can and should still call that practical certainty since for all practical considerations, we are certain enough, even if we are wrong, strictly speaking.
This does by no means address the epistemic problem that we cannot know the actual probabilities of p and whether the difference would change our decision and, by extension, we cannot know whether and at which point we maybe already are, indeed, practically certain, ie. this notion of practical certainty has no bearing on the practical problems faced in decision processes. Therefore, ironically, this notion is absolutely impractical because we are uncertain in our decisions only if our epistemic uncertainty is such that it could make a difference for our decision if actual probabilities differ.
In other words: Well, duh, if we are uncertain in our decisions we are practically uncertain because we are epistemically uncertain. But we also could be practically certain (in Locke's sense) without knowing it, so it actually boils down to an epistemic problem no matter from which angle we look at it. And that is exactly the point Locke wants to stress in that paper.