If your theory were true --- that apparent engagement increases really just represent navigational inefficiency --- you'd see drops in other metrics, like user retention and visits. But you don't. What you see is that people use the thing more. It's hard not to interpret that as better.