I have some issues with the experiment design.
Although they don't explicitly state this, it seems that all of this happens over a relatively short period of time. I suspect that when people normally resort to physical release to blow off steam, they isolate themselves from the person or source of their stress. This may not even be measuring how people typically use this type of behavior to manage their anger.
More importantly, the following comes from the description of the study procedure:
If the participant did not rank the punching bag activity first, the experimenter asked if the participant would be willing to hit the punching bag, explaining that ratings were needed for each activity on the list and that more ratings were needed for the punching bag activity.
Participants were told that they should think about their partner while hitting the bag.
Next, participants completed a mood form that measured anger and positive affect. The anger measure consisted of 15 adjectives (e.g., angry, annoyed, furious) from the hostility subscale of the revised Multiple Affect Adjective Checklist (Zuckerman & Lubin, 1985).
So first of all they had researchers ask people to engage in the activity even if they hadn't originally wanted to, which biases the results. The Milgram experiment suggests that people will go along with what a research asks of them even if it's objectively horrible.
Next, they told people to specifically focus on the individual that had angered them while doing this. I have no idea if it is standard for people using physical activity or even what could be considered violent activity to focus and maintain focus on what has antagonized them while engaging in this activity, but they've introduced another potential source of bias.
Finally, they showed them a list of words that are all associated with anger. This is invariably going to result in priming issues (i.e., would the results be different if they showed some group a list with words like "calm", "patience", etc.?) or possibly influence the mood of the subjects by keeping them focused on how mad they are.
This sounds like far less of a test on catharsis and more likely proof that if you get a person riled up and keep reminding them of their anger that they're going to react more angrily. It doesn't sound at all like they measured catharsis but rather the effects of making someone angry and then keeping them focused on that anger.