Percent signal change is thought to represent, approximately, the amount of the BOLD effect. If the intrinsic EPI data is 20% less intense in the occipital lobe than the frontal lobe (for example, say due to RF coil nonuniformity), then the amplitude of the signal changes will be 20% less also, all other things being equal. Then the fit coefficient would be smaller, but the percent signal change would be the same.
Thus, it is reasonable to convert to % change when you are formally or informally comparing activations acquired at different locations or in different subjects. Also, it is customary and will be expected by the referees of your paper.
bob cox
