It looks like the 18 animals were actually sub-divided into at least three test groups. One test group got the drug after 3 days after exposure, one at 4 days and one at 5 days. So the number of animals in each sub-group is closer to the control.

It wouldn't surprise me if there were actually six test groups of three animals each (perhaps divided by both day-of-treatment and something else like drug amount), thus matching the control. But the article doesn't actually imply that.