If you have an absolute standard curve generated from a known number of target molecules per each standard dilution, and the target used is a single copy gene in the bacterial genome that you are investigating, then 1 cfu would be interpretable in terms of Cq value and the number of cells indicated per cfu.  Knowing how many copies of target per organismal genome would be absolutely necessary as would be generating a stealth absolute standard curve that is diluted in a matrix-interfered buffer that is similar to what your sample extracts would be in preceding qPCR. Further, if your standards are being generated from a known number of cfu's, then the efficiency of extraction would be something you will want to be aware of.  How are you calculating your cfu values in the first place?  By a turbidity test of some kind? (E.g., McFarland turbidity test?).  These kind of estimates of bacterial numbers/cfu have a large error associated with them from the outset.  What method are you using to determine how many bacterium per cfu in the first place?

In the simplest sense:  Initial # of target copies in a qPCReaction = EAMP(b-Cq)

Or:    Xo = EAMP(b-Cq)

Thus:

LOG10{[Xo / (# targets/genome)] / extraction efficiency} = log # of cells indicated in the qPCR reaction

Lastly:  if you find that the above expression tells you that you have 10000 cells worth of target indicated in the qPCR, and you know that e.g. 1 cfu contains 1000 cells, then you can also state that the qPCReaction contained an equivalent of 10 cfu's worth of 'bacterial cell material' in it initially.