From the Section 3.1 of this paper
The Dependency Problem. The main downside of interval arithmetic is that the computed bounds may be extremely pessimistic. As an example, consider the simple operation ← 2 −, evaluated on the range ∈ [−1, 1]. Clearly the actual bound on is [−1, 1], but applying the rules above yields a looser bound of [−1, 2] even in this simple example.
When I try to apply Interval Arithmetic rules I get:
$$f([-1, 1]) = [min(-2, 2), max(-2, 2)] + -1 [-1, 1] = [-2, 2] + [-1, 1] = [-3, 3]$$
I do see that this yields a looser bound but I don't understand why it becomes $[-1, 2]$