You have identified a problem that others have discussed with VSO languages and assumptions about constituency (that there are verb phrases of verb+object, excluding the subject). There is no easy way to do this, so there are two options:
1. Some of the early approaches involved calling these languages "non-configurational". In other words, there is something different about VSO languages (and potentially OSV, though very few languages can be clearly identified as such), because there is no way to group OV or VO as a constituent: so these languages just don't have verb phrases: they're not "configurational". Of course this is not a good analysis, so although it seemed necessary, it was not desirable. (Alternatively we could try to work out whether there is some other kind of grouping in these languages, like V+S, then object later. But I don't know that such an approach has been advocated for. On the contrary, it's popular in Generative grammar to assume that all languages have similar structures, at least at this broad level.) If you want to know more about that, look at early work on Celtic and Semitic (mostly Arabic).
2. We can make the assumption that these languages aren't actually VSO in their structures. Instead, we will analyze them as SVO (or SOV) and then apply movement [or possibly scrambling in some cases, but not when it's the default order] to derive the surface orders. Something attracts the verb to a higher position in the tree, so it moves there. That way, the verb phrase actually does not exist in the final order, but it's there in the derivation for all necessary purposes. Whether we should try to bend languages to fit our theories, or whether there is independent motivation to actually analyze all languages as SVO in some sense is another question, and one that is debated. Remember also, though, that even in many SVO languages, there's subject movement to the beginning of the sentence and then the verb may also move up (it does in French, but not in English, based on arguments like adverb placement). So if we don't raise the subject but we do move the verb, we'd get this order. It means that surface word order is not as important as other arguments for constituency, but it makes things difficult. If you want to know more, look at any modern work on Irish (or other Celtic languages, as well as Arabic, etc.). Andrew Carnie's textbook (Syntax) hints at some of this for Irish, and his research on Irish in general might interest you.
Very short answer: don't try to do this if you're just learning about X-bar theory. It's a problem for basic assumptions without making some sort of exception. That's why it isn't mentioned in any detail in most textbooks. In fact, most would gloss over it rather than either confuse the reader with a much more complex analysis or suggest that maybe the theory doesn't work as well universally as it appears to for basic English data.
