They explain how they assessed shoulder strength in the methods section:
While something like this is far from "real world," it does help control a lot of the variables. When they say it does not improve strength, they are referring to this specifically, and are not making a blanket statement, which they state in the discussion here:
They are comparing pre and post intervention results, as that is what they assessed. They are not stating there was no improvements in military press as they did not look at that. They are not stating there were no improvements in anything else - only that there was no improvements in shoulder strength as assessed by the dynamometer. It is also important to note - which they do at the end for a close reading - that they specifically say "our study protocol." They are, again, explicitly not making a blanket statement.
Folks can argue whether or not a dynamometer has any bearing to real world strength. I'm not going to wade into that, just like I'm not going to argue that the swing improves shoulder strength.
I don't think this is actually the case, as you introduce a lot of variables as far as studying - such as - is the increase strength-specific or technique-specific or a blend. For most people in the real world it might not matter, but for a scientific study, it may be important to suss out the differences - and this is why a lot of times you see things like max iso holds or dynamometers, regardless of training method. They are assessing force output and whether it increased or not. And in the context of this particular study, that removes the need from instructing and assessing yet another movement. Folks can argue this too, but there is a reason behind the methodology.