Too many questions at once, typed out in a format that is very hard to read (all run together). That is why the answers are delayed.

1) The mask dataset must be at the same resolution as the functional/EPI dataset to which you wish to apply it. If your functional dataset is already resampled to Talaraich coordinates at the same resolution as the mask (default resolution is 1 mm) then you do not need to resample the mask. Otherwise, you do need to resample the mask, probably using a command like

3dfractionize -template epi+tlrc -input mask+tlrc -prefix mask_resam -clip 0.4 -preserve

2) The Brodmann areas in the Talairach atlas supplied by U Texas are rather outline-y looking. This is what we have. It can be useful to look at the regions as you draw them in the "Render [new]" plugin. If you set up that plugin with the same overlay dataset that you are drawing into, turn on "ShowThru" and "See Overlay" and "See Xhairs" and "DynaDraw", then as you draw new values into the overlay dataset with the ROI plugin, you will also get a volume rendering update. (You cannot draw into the rendering window, though.)

3) The principal purpose of ROIs, to my mind, is to customize the analysis to the individual subjects. Therefore, I think that you should use the Talairach atlas regions only as a starting point for drawing your ROIs, and then you should manually edit the ROIs to match each subject's particular brain shape.

4) I know nothing about MRIcro.