It was an experiment – not exactly a classic design – but still an experiment. In that context 17 participants are not ideal, but workable and typical. As you can imagine, getting people to sleep in a lab for two nights is hard and costs money that often isn’t there.
I wanted to read the actual paper to give some more detailed insight, but the website of the right database is currently down for maintenance. From their abstract I can see that from the outset they excluded everyone with existing sleeping difficulties. That implies that during recruitment and before the experiment started they already asked questions about sleeping patterns (to exclude outliers from the experiment).
I’m willing to bet that they also controlled for other variables, just like you mentioned, for example other media exposure or the previous nights’ sleep. That’s what you usually do.
(While searching for the paper I actually found some other papers about video gaming before sleeping. It seems like quite some found broadly similar results to this one, so in that context the result doesn’t seem super surprising. But note: This was just me browsing around and glancing at a few abstracts, by no means a thorough or even somewhat acceptable literature review.)