ES 6.6.2 - Synonym Filter Not Working as Expected

[axitxan (Akshit Jain)]

Version: 6.6.2
Issue: A word which is not present in synonym file is getting tokenised as a synonym.

Steps to reproduce

Step 1: Create Index
PUT /test1

{
  "settings": {
    "index": {
      "refresh_interval": "5m",
      "number_of_shards": "1",
      "analysis": {
        "filter": {
          "word_delimiter_with_number_filter": {
            "split_on_numerics": "true",
            "generate_word_parts": "true",
            "preserve_original": "false",
            "catenate_words": "true",
            "generate_number_parts": "true",
            "catenate_all": "false",
            "split_on_case_change": "false",
            "type": "word_delimiter",
            "stem_english_possessive": "true"
          },
          "my_misspelled_synonym": {
            "type": "synonym",
            "synonyms": [ "a6x => xyz" ]
          }
        },
        "analyzer": {
          "my_analyzer": {
            "filter": [
              "word_delimiter_with_number_filter",
              "my_misspelled_synonym"
            ],
            "tokenizer": "whitespace"
          }
        }
      },
      "number_of_replicas": "0"
    }
  }
}

Step 2: Run analyse query
POST /test1/_analyze

{ 
    "analyzer": "my_analyzer", 
    "text": "a 6 x"
}

Output (Not expected since "a 6 x" is not present in synonym file):

{
    "tokens": [
        {
            "token": "xyz",
            "start_offset": 0,
            "end_offset": 5,
            "type": "SYNONYM",
            "position": 0
        }
    ]
}

Step 3: Creating the same index and running the same analyse query on es 2.3.4 results in different output.

Output with es 2.3.4 (Expected):

{
    "tokens": [
        {
            "token": "a",
            "start_offset": 0,
            "end_offset": 1,
            "type": "word",
            "position": 0
        },
        {
            "token": "6",
            "start_offset": 2,
            "end_offset": 3,
            "type": "word",
            "position": 1
        },
        {
            "token": "x",
            "start_offset": 4,
            "end_offset": 5,
            "type": "word",
            "position": 2
        }
    ]
}