Are there infrastructure changes that enhance the safety and efficacy of cycling? Of course. Do all of the changes that have been tried fit into this category? Of course not. It is worse than useless to make sweeping generalizations about "cycling infrastructure" as though there is only one flavor.
For example, a protected lane, even grade-separated, may be an excellent implementation if it is wide enough for its anticipated use and doesn't have cross streets or driveways every few hundred meters and if it is implemented in such a way as to solve, or at least mitigate, the intersection conflicts that come with these nonstandard builds. (Such things as no right on red and giving the bike way intersection priority through either signal timing or sensors can go a long way.) However, squeeze one of those into a street full of driveways and 100 meter blocks and you've got a formula for real problems.
It's the same deal with bike lanes. If they are adequately wide, kept clear of debris, aren't placed in the door zone and handle the intersections well (either ending the bike lane 100 feet prior to intersections without right turn lanes or striping them right up to the limit line with the right turn lane "appearing" to the right of them, which clearly designates the bike lane as the thru lane and lets the turning traffic know they must change lanes and yield to the bike lane they are crossing), then they are great. Miss any of those details, and they can be far worse than nothing.
I guess I see us as a long way from the end of this discussion. We need to continue to press our politicians and engineers to get rid of bad, dangerous infrastructure (of all sorts) and replace it with good infrastructure and to not think that labeling the infrastructure as "bike-specific" is the same as making it good.