a potential bug

Hello NONMEM Users, Subjects received IV infusion (study 2), IV injection or SQ injection (study1). In addition, I had to inject a dummy (amt=1) dose into the 3rd compartment, which describes the target. 2 studies were integrated. The time-concentration plots looked good for one study (IV injection or SC infusion) and meaningless/really bad for the other one (IV infusion; population-predicted values were a few-fold different from the observed ones). The data look correct. The rate, dose, and times look fine (rate=amt/infusion_duration; the first measure is taken after the end off infusion). When I set Rate=0 for the second study, the outcome looks perfect for both studies; the objective function decreased a lot. Because I do not feel comfortable using Rate=0, I tried to change the order of the dummy and real doses, set time of the real dose to 1 second, etc. Nothing works. It looks like a NONMEM bug. Thank you, Pavel