The arguments for segregated cycle tracks (physical
barriers) rather than cycle lanes (paint) can be answered
by one question: "Would you let your child cycle here?"
Where I live, when segregated cycle tracks meet T junctions the layout is often like so:
| C |
----------------------------------| a |--------
Cycles ◅ | r | ▷
----------------------------------| s |--------
divider | |
----------------------------------- Δ ---------
Cars
-------------------------------------------------
This means cyclists using the segregated path have to stop to give way to cars at every T junction - whereas cyclists who don't use the segregated path can maintain a constant speed, with joining cars giving way to them.
This leads to a demographic split in cyclist behaviour. For children and cyclists who aren't in a hurry, the segregated path is more friendly - but for 'serious' cyclists and people in a hurry, staying on the road makes more sense. Especially if there's a high density of T junctions, which isn't unusual in some urban areas.
Fortunately, in my country there aren't any laws that force cyclists into bike lanes, so it generally isn't a problem. But any urban planner who wants 100% of cyclists segregated from traffic is going to have a difficult time on their hands :)