Hi George,
I am not sure why you felt confused with the two datasets. If you have one anatomy dataset (smaller one with only one subbrick) while the other is functional dataset (usually multiple subbricks), AFNI would automatically load each of them respectively. Switching anatomy would not have any effect on the functional dataset. If this is the case, it is not a bug. If the problem is about two anatomy datasets, it might be a bug then.
The number of stumili would depend on the experimental design, and theoretically you need to consider and model all conditions involved in the experiment. Less stimuli would involve less number of parameter estimations, leading to more degree of freedom for parameter estimation, but you'd better include all the conditions to reflect the real situation.
Regarding hemodynamic response function, yes, different curves would lead to different estimations. There are basically two kinds of choices: (1) hrf type; (2) parameters. Once the hrf type is determined, some parameters (such as power or delay for Gamma variate) might have effect on fitting while others (i.e., -peak) are not really esential.
Cheers,
Gang