OK.I just referred to this:And sorry I'm not familiar with much Wittgenstein, such as his 'language games', so you'll need to spell out what you mean.
We can only communicate about what our mental experience is (eg my experience of seeing a beetle), because none of us has an objective 'god's eye view' of reality. That's inherent to being a Subject with a first person pov.
Experience is also inherently private - I can't see what you're experiencing seeing. As if your experience is inside a box I can't see into.
Hence we have to communicate to share experience, because we can't mind-meld. But we can never be sure that what we call blue or a beetle is the same for both of us.
If the language works in a practical sense, then it will become accepted regardless. And we can infer that it's down to sufficient similarities in us which enables it to work on a practical level. (This is where your notion of structure applies?).
But the issue arises because experience is inherently private. (Hence the problem of inverted qualia). And that's why we can't be certain that rocks which can't communicate with us have experiential states, or computers which can communicate with us do. Or anyone/thing but ourselves ultimately (the problem of absent qualia, or philospohical zombies).
Does that make sense? If so, the issue is rooted in the inherently private/first person/subjective nature of experience.
However, our ability to coherently communicate suggests there's enough similarities ('structure' ) in how we experience to make that possible. Neural correlation is a different type of approach which suggests that too. But it's not ultimately knowable, because of the root problem - the private nature of experience. So if we both say each of us has a pain in our left little toe, we can go for neural scans and similar parts of our brain will show similar activity. But ultimately we can't know if we're experiencing something similar, because what we both call pain (or beetles) might be different. What you call pain might be my tickly, or sneezy, or maybe even blue or tree, or it might be a dream, or you might be a philosophical zombie, or might not exist, nothing might exist but 'my' experiential states. Everything else is inferred, based on the content of those experiences.
The only thing which can be known for certain are 'my' experiential states exist. And their properties are their content. I can only know you exist as a content-property of 'my' experiential states. I can only know my own body and brain exist in the same way.
That's why any hypothesis which says experiential states aren't real or only relational doesn't work. It's the public/shared act of communicating (language) which is relational, not the experiential states themselves. .