I have documents each of which have many nested objects containing analyzed text (think paragraphs of a document). I am running nested span near queries in order to find the documents containing phrases and defining inner hits on the query so I can identify which nested objects matched for each document.
I get inner hits, and in general when there are multiple matching nested objects I get multiple inner hits for a document - but often I am missing some. E.g. there may be 5 nested objects in a document which should match (by inspecting the text), but I only get 3 inner hits for the document.
Is this behavior known? Is it intentional for some reason? The only plausible reason I can think of for it being intentional is that the excluded hits are lower-scoring, although they should not be in the cases I have checked (and I definitely see cases where I get back inner hits for a document with different scores).