It is my understanding (or perhaps conjecture?) that the failure point of the rear towers is not the top, flatter part. Rather, it lies in the stresses and hardening that is developed in the curved part that has been stretched into shape originally, and has transferred impact loads for 40 some years since. To Rob's point, the milled cap solution replaces this stressed section and connects down lower to the cylindrical section where the load is almost purely tensile and in the direction of the metal - thus much less likely to be fatigued like the top (regardless of whether it is glued or welded - that is another debate). I like the proposed advantages of gluing if the parts fit up nicely. But if the car is completely apart (bare chassis) and not painted yet (I am thinking of my car eventually), I'd likely weld a solution in.