Activation energies of defect migration in ABX3 perovskite halides are widely obtained through static supercell calculations with the nudged-elastic-band method. Taking methylammonium lead iodide (CH3NH3PbI3, MAPbI3) as an example, we demonstrate that such calculations are unreliable for the higher symmetry structures adopted by the material at temperatures relevant to device operation (tetragonal and cubic MAPbI3) because, in addition to ion relaxation around the point defects, local structural modifications characteristic of the ground-state (orthorhombic) structure occur. In this way, we offer a simple explanation of why calculated activation energies of defect migration in MAPbI3 suffer from surprisingly large scatter. We propose a robust test to determine whether static supercell calculations of point-defect processes in ABX3 perovskite systems are reliable.