C,
I can save you quite a bit of time and effort. Instead of using two separate indicators, you can use one indicator with two states. On slide 1.12, there is an example of what I mean. For Line 1, I copied the check and deleted it. Then I edited the states of the X. I duplicated Normal and named it Wrong, then duplicated it again and called it Right. In the right state, I deleted the X and pasted the check. Now it has a Normal state (my experience shows me it is best not to use Normal state in conditions, unless you are very careful.), a Wrong State, and a Right state, with its initial state set to Hidden. What that does for you is to cut the number of triggers in half when Submit is clicked; you only need to change the state of one object depending on the value of the text entry. The two triggers that set the states of Correct and Try Again need to be changed to reflect the new state names. (Just my personal oddity; I don't use names that the system uses, like Correct and Try Again. It helps me know if I'm dealing with my objects, or the system's objects.)
All of that is for free. To answer your question, I think the Try Again button has some corruption. That happens a lot more often that we wish it would. When I deleted it and recreated it, it began to work correctly.
I also took off the Try Again trigger that jumps to the same slide. That really interferes the record keeping that is built into tracking the number of attempts. Just hide the Try Again layer, and the learner can make changes.