Yep margin of error could go either way. Could have gone the other. Thats why controlled trials are run to cut down on the variables and then rerun if there is a signal to be found. Trial failed based on top line data for sure. Whether there was a different signal on complete data we won't know for awhile and may not matter anyway
I assume your error margin is an INDIVIDUAL error margin. The GROUP as a whole should be closer to what would be expected for any given trial in moderate-to-severe AD patients. That is, trending downward over time.