I understand that, do failure analysis on a fairly regular basis concerning support structures (automotive door pillar beams and suspension). Thing is we do the stress test, get the nice colored renderings showing likely failure points - then I have it machined/formed and do a "real-world" test and it's different. You cannot account for everything because we don't understand everything.
Edit: Again, yet to see/hear of any rear engine mounts suffering catastrophic failures in the "real world". No matter how much analysis is put forth, until something is physically proven - it remains a theoretical outcome.