Multi_match search appears to be using the index analyzer

[Cara]

I'm using Elasticsearch 5.5 on AWS.

With a type (with NEST mapping attributes) that looks like this:

	[ElasticsearchType]
	public class MyType
	{
		public const string AutocompleteAnalyzer = "autocomplete";
		public const string StopWordsAnalyzer = "stopwords";

		public string Id { get; set; }

		[Text(Analyzer = AutocompleteAnalyzer, SearchAnalyzer = StopWordsAnalyzer)]
		public string Name { get; set; }

		[Keyword]
		public string NotAnalyzedProperty { get; set; }

		[Text(Analyzer = AutocompleteAnalyzer, SearchAnalyzer = StopWordsAnalyzer)]
		public string OtherTextProperty { get; set; }
    }

The index is created like this:

var indexDescriptor = new CreateIndexDescriptor(IndexName)
    .Settings(s => s
        .Analysis(a => a
            .TokenFilters(t => t.EdgeNGram("autocomplete_filter", e => e.MinGram(3).MaxGram(20)))
            .Analyzers(aa => aa
                .Custom(MyType.AutocompleteAnalyzer, c => c.Tokenizer("standard").Filters("lowercase", "autocomplete_filter"))
                .Standard(MyType.StopWordsAnalyzer, w => w.StopWords(_stopWords)))))
        .Mappings(ms => ms.Map<MyType>(m => m.AutoMap().Properties(p => p.Text(f => f.Name(t => t.Name).Fields(ff => ff
            .Text(tt => tt
                .Name(AutocompleteField)
                .Analyzer(MyType.AutocompleteAnalyzer))
            .Keyword(k => k
                .Name(KeywordField)
                .IgnoreAbove(256)))))));

var createIndexResponse = await _elasticClient.CreateIndexAsync(indexDescriptor, cancellationToken);

When I run this search:

{
    "size":10,
    "sort":[
        {"_score":{"order":"desc"}},
        {"name.keyword":{"order":"asc"}},
        {"id.keyword":{"order":"asc"}}],
    "query":{
        "bool":{
            "must":[{
                "multi_match":{
                    "type":"best_fields",
                    "query":"raining",
                    "fields":[
                        "name.autocomplete",
                        "otherTextProperty"]}}],
            "filter":[{
                "bool":{
                    "must":[{
                        "match":{
                            "notAnalyzedProperty":{"query":"Test"}}}]}}]}}}

It matches any document that contains any word starting with "rai", instead of just those that contain "raining". If I add "analyzer":"stopwords" to the multi_match query, it works as expected, and only finds documents containing "raining".

My understanding from the docs, and also other questions (https://stackoverflow.com/questions/36051750/elasticsearch-multi-match-cross-fields-query-with-different-query-analyzers), is that the search should default to the search analyzer defined in the field mappings, but it appears to be using the index analyzer instead.

I assume there is something wrong with my mapping, but I can't see what. Any ideas?