Highlight parameter 'boundary_chars' does not work as expected

Elastic Stack Elasticsearch
1

I used hightlight with parameter 'boundary_chars', but not always get fragment truncated at boundary characters.

Example:

PUT wpz_2

PUT wpz_2/_mapping/test
{
  "properties": {
    "test": {
      "analyzer": "index_ansj", 
      "type": "string", 
      "term_vector": "with_positions_offsets"
    }
  }
}

PUT wpz_2/test/3
{
  "test": "，全市工业80%以上的大型装备实现了信息化集成。投资2000万元启动“智慧企业”专项行动，重点支持工业企业无线、物联技术应用。"
}
GET  wpz_2/test/_search
{
  "query": {
    "match": {
      "test": "智慧"
    }
  }
  , "highlight": {
    "boundary_chars": ".,!?；，。？！",
    "fragment_size": 30,
    "fields": {
      "test": {}
    }
  }
}

Output:

{
  "took": 4,
  "timed_out": false,
  "_shards": {
    "total": 5,
    "successful": 5,
    "failed": 0
  },
  "hits": {
    "total": 1,
    "max_score": 0.057534903,
    "hits": [
      {
        "_index": "wpz_2",
        "_type": "test",
        "_id": "3",
        "_score": 0.057534903,
        "_source": {
          "test": "，全市工业80%以上的大型装备实现了信息化集成。投资2000万元启动“智慧企业”专项行动，重点支持工业企业无线、物联技术应用。"
        },
        "highlight": {
          "test": [
            "集成。投资2000万元启动“<em>智慧</em>企业”专项行动，重点支持工业企业无线、物联技术应用"
          ]
        }
      }
    ]
  }
}

The highlight fragment is not truncated at '。' (third character)
I also tried standard analyzer, result is the same. Can I always get fragment truncated at boundary characters? Here, what I expect is:
"投资2000万元启动“<em>智慧</em>企业”专项行动，重点支持工业企业无线、物联技术应用"

I'm using ES 2.3.3

Best regards,
Abel