Interesting research but very early stages. There's a good chance that this isn't really specific to bikes, but more generally applies to the benefits of people enjoying hobbies and a good quality of life translating into a better psychological state, which is known to affect health outcomes in a myriad of ways specially related to brain function and hormones.
Interestingly, the data also shows that mixed-walking is associated with reduced dementia as well... and I find it a bit confusing that the authors are lumping driving and public transit together in the nonactive category because public transit usually involves a good deal of walking too without people categorizing it as a walking mode of transport.
Beyond separating public transit from driving, I think major factors to control for are: 1) how often these non-commute activities happen; and 2) what's the stress level of that transportation mode. With those three things in mind I think it would clarify why would walking be associated with worst outcomes than nonactive...