Usually (in over 95% of cases) you'll want to set bounding box (bbox) first, and then use fit-to-content-width zoom mode. In that mode, page forward/backward will move viewport, and when the end of the page is reached, it will change page. If page mode is selected, page buttons will always change page.
Page/viewport mode switch only works (has effect) in "smart" zoom modes - "smart" zoom modes being fit-to-content-width and 2-column-mode.
The largest part of the remaining 5% of use cases, user will want to use 2-column mode - usefull for reading texts with 2 columns. In 2 column mode, pressing page forward/backwards (as well as pressing fiveway keys) will move viewport so the next part you need to read comes next automatically.
In all other modes, pressing fiveway just pans the page predetermined amount of pixels (very rarely useful).
I'd say that of the remaining zoom modes, the only one that's somewhat useful is full page mode, for setting bbox. The rest of the modes - well, maybe in some very, very rare occasion
I hope you'll find this at least a bit helpful.