On the transition times in a loss model of wireless networks