First of all, thank you for your prompt response and valuable suggestions.
1) Regarding the source locations in MNI I didn't quite understand why this doesn't matter. The DCM model requires an EEG signal after source reconstruction, right? So the space will be the source space instead of the sensor space (the actual electrodes), right? If so, how come the resolution of the coordinates doesn't make a difference? Isn't it possible that sources are several mm misaligned with corresponding locations in MNI ?
2) About data epoching: I have a continuous signal acquired during performance of a task constituted by 4 conditions: 8 periods of "baseline" (22 seconds), 5 periods of "condition A" (18 seconds), 4 periods of "condition B" (18 seconds) and 3 periods of "condition C" (18 seconds). I was thinking about separating my continuous signal into epochs of equal length to the periods of each condition. So, for example for "condition A" I would have 5 epochs of 18 seconds each corresponding to "condition A " which would then be averaged into one single epoch. Does this make sense?