Based on radar?EF5?
This is a fascinating case. I have a tornado intensity forecasting index (the Vortex Value) which, even upon detailed reanalysis (e.g., VWP, sfc conditions etc.), failed to to indicate greater than minimal EF3 for the Newnan area. Must have been an extremely localized area of EF4 damage, perhaps enhanced by terrain flow or another oddity. Correlation coefficient data for the vortex value is over 0.9.