In practice, it's hard to notice any advantage smaller than "one-third of a Pawn", but it's possible. What's more, you can notice differences of advantage between one-third and two-thirds.
I would go further and say that the main reason _Point Count Chess_ is unusable in practical play at the master level is because not all tempi are of equal value, not all doubled Pawns are equally weak, and so on. In an unbalanced position where each side has five or six "points", the cumulative error is often more than one "point".
However, the concept of the point as a quantum of advantage is useful both in theory and practice; at least if you're not a Grandmaster, it's hard to detect much of a difference between an advantage of "one-third of a Pawn" and an advantage of half a Pawn, but the difference between a one-point advantage and a two-point advantage is something you are sure to feel.
For the theory of the values of chess pieces, the significance of this is that a point is more than ten percent of the value of a Knight. Since there are two Knights, if you want to replace one side's Knights with a piece of equal value, it would seem that you need to find a piece whose value is at least 0.95 times that of the Knight, and no more than 1.05 times the Knight's value.
I talked about a one-point advantage, but how much is that in normal terms?