Preprocessing in CONN optional. If your data are already preprocessed than you're in the clear. As long as you have information like framewise displacement and head motion parameters, you can make your own first-level covariates (e.g. scrubbing, motion). For denoising, you can just move all the nuissance regressors to the right side, turn off all filtering and other options, and "run" denoising. However, unlesss you already scrubbed your data to censor motion, you would need to denoise. Does this make sense?

Best,
Steven