The reason for this is that mathematical papers are fundamentally about an insight about mathematical objects. The insight typically comes from the mathematician having a deep understanding of the mathematical objects. The theorems are some formalised statement that should be true based on the deep understanding. The proofs are something else entirely, a sequence of steps which combine to verify the theorem but where many steps may result from trial-and-error searching for a path forward. Coming from a good mathemematician, the original insights are much more likely to be correct than the details of the proof steps.