as You already said, MT Spring is FK only. If you want 'true' pinned simulation, solver has to support that. AFAIK there is nothing in MT Spring code, that looks like this.
Now, you can do some 'post-process', let's say, use a two independent MT spring chains, one starting from root bone, another would be reverse, starting from effector. Then you constraint bones in the 'final' chain, gradually from root to effector, to both MT spring chains. But all that has some sense only on chains with much more than two bones. And there is a good chance for unwanted effects. At least, won't be easy to keep the effector 'absolutely' pinned.