Unexpected hits when using n-gram

Hi, I am using n-gram filter with my index.

Index configuration:

PUT dev-dk-ngram-test
{
  "settings": {
    "index": {
      "max_ngram_diff": 2
    },
    "analysis": {
      "analyzer": {
        "default": {
          "tokenizer": "whitespace",
          "filter": [ "lowercase", "custom_grams" ]
        }, 
        "default_search": {
          "tokenizer": "whitespace",
          "filter": [ "lowercase", "custom_grams" ]
        }
      },
      "filter": {
        "custom_grams": {
          "type": "ngram",
          "min_gram": 3,
          "max_gram":  5
        }
      }
    }
  },
  "mappings": {
    "properties": {
      "dk-ngram-test": {
        "type": "nested", 
        "include_in_parent": true, 
        "properties": {
          "name": { 
            "type": "text"
          },
          "desc": { 
            "type": "text"
          },
          "age": { "type": "integer" }
        }
      }
    }
  }
}

Them I am adding few records:

PUT dev-dk-ngram-test/_doc/8?refresh
{
  "dk-ngram-test": {
    "name": "Atlas",
    "desc": "Domain",
    "age": 1
  }
}

PUT dev-dk-ngram-test/_doc/9?refresh
{
  "dk-ngram-test": {
    "name": "Functional Classification",
    "desc": "Functional Classification",
    "age": 1
  }
}

Now I am searching "Atlas".

GET dev-dk-ngram-test/_search
{
  "query": {
    "query_string": {
      "query": "atlas",
      "escape": true
    }
  }
}

I expect the Document with ID 8 to be of higher sore, but instead the document with ID 9 has higher score.

To give more background, we started with default (standard) settings, but it didn't give us results which had a minor spelling mistakes. So we switched to n-gram. It works well for our most search use cases, but searches like above are causing issue. The term which match exactly are getting lower score at times. Could you please help.