The sequence of events is:

• editCheck-addReference : action-reject - the user chose "no"
• editCheck-addReference : edit-check-feedback-shown - the feedback form was shown
• editCheck-addReference : edit-check-feedback-reason-[reason] - the user chose reason in the form

Notably, there's a "back" button on the feedback form. The user can click it rather than choosing an option -- it won't dismiss the check, but it'll let you go back and either pick another option or go even further back and edit the content through normal means. However, it's not instrumented when people choose it.

So: this 11% is either people who have gone back to choose another option, or who have given up there-and-then and closed the window.

So: this 11% is either people who have gone back to choose another option, or who have given up there-and-then and closed the window.

To be doubly sure, I'm following...

When you say, "people who have gone back to choose another option..." are you referring to people who follow the following sequence?

1. See Reference Check
2. Decline to add a reference
3. See the feedback form
4. Tap Back from within the feedback form
5. Elect to add a reference

Yes, though as well as "5. add a reference" they could also have either removed the triggering content from their edit, or abandoned their edit entirely.

(Analysis of the edit session would let you broadly determine which of those was the case, though it'd be pretty in-the-weeds, and you may not be able to tell the difference between removing the content and adding the reference manually if the user was being served multiple reference checks.)