It's wrong for a researcher to do such a thing because of the bias or skew that may be created or not accounted for.
For example, this skew could be best seen in an experiment where an individual's socioeconomic status plays a key role. In this experiment, the first group may be on time because they all have cars and are able to transport themselves along with being able to make flexible arrangements allowing them to arrive on time. On the other hand, the second group may consist of individuals who were forced to take the bus because they don't have a car due to its prohibitive cost. In addition, they may be forced to work on the hour causing them to be unable to make flexible arrangements. Therefore, we see that the first group may consist of individuals with a generally higher socioeconomic status while the second group consists of those at a lower socioeconomic class.
This is just one example, but the key is that it would be biased and would fail to account for an uncontrolled condition.