Question: Completion suggester behavior on String with number

I've build an Auto-complete service with Elasticsearch 2.x.x. and one of my auto-completable values is "3M". I've configured the fuzziness to AUTO and my mapping is just default:

"mapping": {
    "type": "completion",
     "analyzer": "simple",
     "payloads": false,
     "preserve_separators": true,
     "preserve_position_increments": true,
     "max_input_length": 50

Based on this documentation the analyzers should be simple and Fuzziness AUTO means at MAX 2 spelling errors when a auto-completable word is >5 characters.

Here's the problem, whenever I type "1000000M" it still auto-completes "3M" although 1000000 and 3 are exceeding the limit of 2 spelling errors.

Does Elastic knows 1000000 and 3 are both numbers and I'm looking for {a number}M?

I would like the numbers to behave as actual String spelling errors because this is not the preferred behavior.

Even when I set Fuzziness to ZERO, it still corrects 1000000M to 3M.

original post: