Bicycle Detection: Inductive Loops or Push Buttons? What do Cyclists Choose

We have an approach to the signal at NE 7th & Broadway (the southbound direction) where there is an exclusive right turn lane and a through movement (no left turn onto the one way street). We have placed a detector in the through lane and have a bicycle push button. There was a complaint that the sign for the bike button was resulting in people on bikes leaving the through lane to go push the button. The logic was that the sign was directing people to use the button.

Our intern observed the intersection on two separate occasions and found that 42 cyclists had to stop at the intersection to wait to cross Broadway.
79% waited in the thru lane
21% waited in the right turn only lane.
71% of riders activated the light using the pavement marker over the loop
17% used the button
2% used both
10% used neither (cars would activate the green as well).

The riders who did use the button did not seem to have trouble accessing it from the right turn only lane and 80% remained in lane after using the button.

2 out of 42 cyclists violated the signal and went through on red. The average delay was 31 seconds, which would meet standards in the Dutch CROW Manual.

We can do better though and we're exploring whether we can provide a confirmation light on the push button, so if someone on a bicycle is on the detection in the lane, they would know they were detected by the LED being illuminated in the push button. We are also proposing to change or eliminate the sign because it would seem intuitive to most users.