What do you mean when you write nanoparticle? I guess from the pictures, it's a cut-out piece of crystal. Then it is normal that after relaxation the atoms are strongly shifted relative to each other when compared to their positions in the crystal. If we do something like simulated annealing, we can get even more different structure.
I think you are optimizing a structure of CsPbI3 perovskite along with an amine molecule. I don't know what you're up to but there is a rule of thumb saying that in most of the time if the structure changes are too much you are on the wrong way even if the optimization gets complete. Try optimizing one unit cell of the perovskite and then see the changes.