I agree. The 5e rules say moving between attacks is allowed. That is the general rule. The general rule on taking bonus actions is, "You choose when to take a bonus action during your turn, unless the bonus action’s timing is specified." The timing of the GWM bonus action is not specified in the feat description. The prerequisite that must be met to get the bonus is specified, but that is the nature of all bonus actions. The general rule still applies.
A specific exception to the general rule requires an explicit statement to that effect. In this case, there is no exception explicitly stated.
In my experience, a strict literal interpretation of the rules text is the correct one. Every once in a great while, Jeremy Crawford says otherwise, overruling RAW in deference to his original intent. But I've only actually seen this happen once, and I follow his feed on Twitter, and I query Sage Advice via Google searches rather assiduously.
In the great majority of cases, questions posed to Crawford reveal what the petitioner thinks the rule "should" be, rather than what it plainly states, or fails to state. In almost every instance, they are rebuffed. And were I to see this bonus attack issue brought up on Crawford's twitter feed, I would expect him to affirm your position, and I would categorize the petitioner as another example of someone who has reasoned themselves to what the rule "ought to be," in the absence of any actual supporting text.
The one counter example that I have run across is the ruling that the Resilient feat can only be taken once. It does not say this in the text, or in the published errata, but when asked, Jeremy Crawford stated that such was the case.
Now, one might argue that the line in the Elemental Adept feat that states, "You can select this feat multiple times," by inference indicates that all feats lacking such a statement can only be taken once. And if you did infer that, you would have been right. So maybe this is not the cleanest example of textual literalism overturned. But to get to the correct answer about the Resilient feat did require the reader to infer a general rule about feats from a specific rule applied to a single feat. And that's not common in the 5e rules. Rather it is exceeding rare, maybe even unique. And contrary to this example, it is inference, in my experience, which most often leads people astray. To get it right in almost every instance, infer as little as need be inferred.
Not to get all Talmudic.
