Follow up to previous thread re: logisitc regression

This is an interpretation question. I have run a model with a binary dependent variable and both binary and non binary independent variables.

For discussion's sake, lets say that my dependent variable is the passing of a class. 0 if a student fails, 1 if they pass. Among the variables, one is "attended study hall at least once" where 0 is if they did not and 1 is if they attended at least once. The coefficient is .67683 and is statistically significant. I need to articulate how this improves the probability of passing the class when all else is held constant and of course, to do that, I need to understand it myself and I don't.

I took e ^ .67683 and got a value of 1.96763. When I subtract 1, I obviously get .96763. Am I to understand that this is telling me that students who attend study hall are 96.8% more likely to pass than those who don't?
2 follow ups, and maybe its that im so used to linear regression.

In my previous example, I did .96763/1.96763 and got .4917......Is it accurate to say "those that attended at least 1 session in study hall are 49.2% more like to pass the class than those who did not?" (I am not going to include the 95% CI for my report to this person).

Upshot of that, lets say i break it out and have the number of times a student attended, from 0-5. I get a coefficient of .44. Then e ^ .44 = 1.552707, then .552707/1.552707 = .355964. How would I convey this probability in terms of a sliding scale? "Each time a student attends a study hall session, the odds of passing the class rise by 36%?" that doesnt make sense to me, so i assume thats clearly wrong.

Any help appreciated.