Freetext Field Multi Match Query Inaccurate Results

Question

I have an API that builds the following query to Elasticsearch:

   "should": [
        {
          "multi_match": {
            "query": "1225",
            "fields": [
              "id^1.0",
              "title^1.0",
              "titleNgram^1.0"
            ],
            "type": "most_fields",
            "operator": "OR",
            "slop": 0,
            "prefix_length": 0,
            "max_expansions": 50,
            "zero_terms_query": "NONE",
            "auto_generate_synonyms_phrase_query": true,
            "fuzzy_transpositions": true,
            "boost": 1
          }
        }
      ]

id field: client-built. No distinct pattern or length, can be a mix of letters and numbers, sometimes filtered by prefix, cannot be reliably programmatically identified as an ID from the freetext search query.

title: string search

Problem: The ngram field is causing many irrelevant results to be returned. E.g. query:"nurse" results: "nurse", "customer service representative"

Removing the ngram field from the query helps, but then my search is not quite fuzzy enough to differentiate between possibly relevant terms. E.g.:

query: "Police man" results: "police man, milk man, pizza man" missing from results: "policeman"

Are there any suggestions for what I can tinker with here get more accurate results?