Sometimes you have to enlarge the cut so that it protrudes well from the part/assembly to be cut or, as gt22 says, there is a surface part that causes the error.
Then, if this solves the problem and to find out which parameter was the cause, you can (but it's tedious!) extract the two registries (the one that works and the one that doesn't) and compare them with WinMerge for example. But it takes time!
Hello, the same thing happened to me on an assembly plan with the same cut as the one on the detail plan of the part concerned. It turns out that on the detail plane the cut line cuts the 3D correctly, while on the overall plan it is not, and it puts itself at fault.
And as surprising as it may seem, it turns out that it was simply related to a 180° offset position of this part on its axis of rotation, and yet it is symmetrical, a simple reversal of stress on coincident planes can save a lot of time... (but why ??? I'm still looking...)