Hi,

I am running into a similar issue and have a similar question as the original poster that did not get resolved. 
I imported data into 20.b from fmriprep that contained 2 runs of separate tasks (and therefore different lengths and different sets of conditions). CONN loaded these as two sessions, and when I attempted to extract time series for a given ROI, the resulting time series was the concatenated length of the two separate sessions. 

Can anyone confirm if denoising was done SEPERATELY for the two sessions, and then the resulting time series were concatenated?
Also, if I move forward with 1st and 2nd level analyses in Conn, are there any downstream issues with these sessions of different task runs being combined into one? 
Thank you for any help!

Matt