Hello Josh,
Well, the design may be sub-optimal in that the response to cigarette is considerably correlated with that of the neutral cues. You should examine your design matrix to find out how correlated the two stimulus response models are. The more correlated two responses are the more data one needs to tell them apart. The program ExamineXmat can help you check on the amount of correlation present between your regressors. Also, afni_proc.py's review scripts provide some information on correlations between regressors.
cheers,
Ziad