Not sure how well this will work but if you consider the control to be in a block of it's own (N=0, B=6), you could try:
proc mixed data=yourdatainlongform;
class treatment block rep;
model value=treatment;
random block treatment*rep;
lsmeans treatment/diff=control;
run;
Key to this is getting your data into a format where each record has unique values for value, treatment, block and rep. This works well on a simulated data set.
SteveDenham