Unless I am mistaken in order to run a direct analysis, you specify notional loads. However, per the specification it is allowable to use direct modeling of imperfections per C.2a. There is a way to specify camber and drift for members, but from what I can tell it looks like those only apply to the Imperfection analysis and not to the direct analysis.
This is a sound idea. As you may know we do provide a method for performing an analysis using imperfections (sway/camber) but using a generic analysis method rather than using the modified tau method defined by the direct analysis method. However, this would seem like a reasonable request and will be added to the development backlog.