Applying boosts inside a SpanQuery - solr

Question

Is there a way to apply a boost inside a SpanOrQuery? (Solr 7.1, lucene of same version.)
Example of query structure generated by edismax query parser (inside the usual +BooleanQuery / DisjunctionMaxQuery for multiple fields):
SpanNearQuery
List of SpanOrQuery
List of SpanTermQuery or SpanNearQuery <-- I want to boost those terms
I want to apply a boost to each clause inside the SpanOrQuery. So I tried (via custom query parser extending edismax):
SpanNearQuery
List of SpanOrQuery
List of SpanBoostQuery, each wrapping:
a SpanTermQuery or a SpanNearQuery.
The boost seems to be ignored, although the query is executed successfully (except boost not applied).
Here is the use case:
The input is a sentence, a phrase query (quoted with slop) to valuate proximity. I am using edismax query parser and a SynonymGraphFilter and I want to apply a different boost for each synonym, so the boost information is attached to each term in the synonym file (e.g. 0.7_foo). In case of multi-term synonyms (expanded version), edismax generates a SpanQuery, which sounds to me like a graph search, so far so good.
Where it fails is when I insert a SpanBoostQuery to wrap each clause inside the SpanOrQuery (via a custom query parser extending edismax, extracting boosts from term text). While the query is still returning results, the boost seems just ignored.
Is that a misusage? A bug? Any advice about how I can fix it or work around it please?
To have a similar behavior, I replaced the SpanQuery by a list of PhraseQueries, one for each possible combination of terms in the SpanQuery, which can result in lots of phrases, the performances seem highly affected and we lose lots of information.
Thanks a lot!
Edit:
The tree above describes the query generated by my custom query parser (I checked the resulting Query by debugging it), but for whoever wants to figure out how this is implemented, I first let edismax return a parsed query, then I go through the tree recursively, rebuilding a new Query, until I reach the leaf (SpanTermQuery), and I apply a code like below to wrap the term in a boost (SpanBoostQuery is provided by lucene).
if (initialQuery instanceof SpanTermQuery) {
SpanTermQuery q = (SpanTermQuery) initialQuery;
// parse and extract term/boost from q.getTerm(),
// e.g. "0.7_foo" -> {term: foo, boost: 0.7}
q = new SpanTermQuery(term);
if (boost >= 0 && boost != 1) {
return new SpanBoostQuery(q, boost);
} else {
return q;
}
}
I am aware it is not an optimal solution, but so far it works except the missing boost. Advices are welcome, but it would be a separate topic, and not as important for me as my question above :)