Logically, we could parse it as either:
[Craig] [posted a question] [to Larry]
Where "to Larry" modifies the verb phrase "posted a question".
Or as:
[Craig] [posted] [a question to Larry]
Where "a question to Larry" is the object of the verb posted.
We can demonstrate that both work logically, by considering:
"Craig, did you post that question to anyone?"
"Yes, to Larry."
And:
"Craig, did you do anything about that question to Larry?"
"Yes, I posted it."
And of course, they amount to the same thing.
Now, as far as I can see, the grammar of the sentence matches the logic described above. While they're different, they amount to the same thing in terms of meaning. A difference, that makes no difference, is not a difference.
We could say explicitly match the first with "Craig posted a question, to Larry". We can't make something match only the second and not the first. This would be irritating if we needed to make a Reed-Kellogg diagram of it, but that's the Reed-Kellogg diagramming system's problem, not ours. (Reed Kellogg was designed to model the parsing of sentences in English, and potentially other languages. English was not designed to be modelled by Reed Kellogg).