Any csy other than WCS cannot find reactions. I think there are some ways to work around, like Skunks demonstrated.
Make WCS = Desired CSY (Cartesian only)
One way is to assemble everything in a new assembly where the alternate coordinate frame (CS17) is assembled to the default coordinate of the new assembly file. Thus the WCS is the same as CS17.
Another way would be to transform the entire part (within itself) to align to the WCS like it did to the CS17.
To get a reaction measure the constraint must be defined in WCS, but the measure does not have to be as it can transform the reaction to another coordinate system. This gives the reason why you can choose other coordinate systems with your reaction measure. This might work for enforced displacements if there are no free directions... (have to enforce all 3 translational d.o.f.) Make the enforced constraint by defining the 3 components in the WCS, then the measure can have a user CSY to resolve the force in a certain direction.
If you are enforcing in X but leaving Y, and Z free, I would go with my first method of making WCS = Desired CSY.
More on this subject...
You can also use the common engineering practice of slicing/sectioning. This uses an internal cross section (purple bonded surface) created by a volume region.
Also if you use a rigid spring rather than a beam (Skunks) you can measure the reaction on the spring directly rather than using the beam result along curve.
When using enforced displacement the reaction can be found at whatever other constraint is on the model that restricts movement in the direction enforced. This means you are not restricted to getting the measure at the constraint that does the enforcement. Another constraint likely sees the same reaction as in my example the left end of the beam is held all dof by the WCS so it can be used to get the reaction, in addition to any slice between this constraint and the enforced displacement. It should also make sense that slice 3 has essentially zero reaction force.
Note: The first reactions I reported below and the last both have 3mm enforced displacement. The reason the last method with the spring has lower reaction reported is because it allows rotation at the small rigid region where the spring attaches whereas the first method does not. That is the small rigid surface region used to remove singularities is also forced to X -3mm over the entire surface. So for the second example only the end point of the spring is forced to the 3mm and the rest of the rigid region can rotate freely which reduces the reaction force and is probably a more correct situation. Take-away. Be careful with enforced displacements, look closely at deformed models of the results.