I'm a bit ambivalent about that feature. I do think many folks will want/expect it, and it would be useful at times. At the same time, the primary use of the app I personally wanted is to add things I wanted to remember permanently, and let the algorithms handle all of the work of scheduling optimal practice. Gonna think about this.
This is not naive bayes and does not assume independent observations on the exercises. The point of using a network is to model the joint distribution with dependencies.
The 'E' in the regression is the inferred/predicted value of the E variable for that exercise, using no problem history from that exercise--only what's pulled in through the Bayes net. (Sorry that wasn't clear)
The 'T' variable is likely just a case of multicollinearity with the 'E' variable and should go away on a full-scale data set. If not it can easily be removed from the model. The 'E' variable is dominating because is additionally captures cross-sectional affects across the various exercises in the regression.