Not really. The inconvenience with mixed models is that are hard to interpret, and their group-level analysis is complex (see Coregistering subjects with mixed models).
@Sylvain, some additional comments?
Regarding parcellations, as the source space is mixed (surface and volume grids, cortex and subcortical structures), it is not possible to have an atlas to contain surface and volume scouts, although this is manageable by using 2 atlases.