Well I thought that sounded a bit strange - hardly a good representation of a 21% success rate - but I decided to try it. I moved a couple of random armies a few squares and tried again. Success first time! Weird calculation apparently.... thanks Grombeard.