Teodora,
If you did 3dDeconvolve with option "-polort 1" or by default, then you need
0 0 1 1 -1 0
for contrast of (c1+c2)-c3. The first two 0's are for the linear baseline model.
For the 2nd question, I am not so sure what those three subbricks are. The result would be just exactly the same entity as those three subbricks with no statistics attached.
Gang