query topN with join the result not rollup

i use topN query with join，result not rollup，only i use groupBy the result will correct. how can i use top with join?
{
"queryType": "topN",
"dataSource": {
"type": "join",
"left": "xxx",
"right": {
"type": "query",
"query": {
"queryType": "scan",
"dataSource":{
"type":"table",
"name":"yyy"
},
"context": {
"timeout": 180000
},
"intervals": "2020-07-02T00Z/2020-07-03T00Z",
"granularity": "all",
"columns": [
"uid",
"channel"
],
"limit": 100000,
"lenientAggregatorMerge": false,
"descending": false
}
},
"rightPrefix": "r.",
"condition": "uid== \"r.uid\"",
"joinType": "INNER"
},
"context": {
"timeout": 180000,
"priority": 2147483647,
"queryId": "wwh-test",
"skipJwoodCache": false,
"weight": 0
},
"intervals": "2020-07-02T00Z/2020-07-03T00Z",
"filter": {
"type": "in",
"dimension": "r.channel",
"values": [
"googleplay",
"appstore"
],
"lowerStrict": false,
"upperStrict": false,
"alphaNumeric": false
},
"aggregations": [
{
"type": "thetaSketch",
"name": "!Theta_counted_uid",
"fieldName": "counted_uid",
"byRow": false,
"inputThetaSketch": false
},
{
"type": "doubleSum",
"name": "item_golds",
"fieldName": "item_golds",
"byRow": false,
"inputThetaSketch": false
}
],
"postAggregations": [
{
"type": "thetaSketchEstimate",
"name": "counted_uid",
"field": {
"type": "fieldAccess",
"fieldName": "!Theta_counted_uid"
}
}
],
"granularity": "all",
"lenientAggregatorMerge": false,
"merge": true,
"dimension": {
"type": "default",
"dimension": "r.channel",
"outputName": "channel"
},
"threshold": 500,
"metric": {
"type": "lexicographic"
},
"descending": false,
"virtualColumns": []
}

Reproduce the supplied TopN query with its join and compare the result with the grouped variant. Start by tracing the TopN and join query paths in Druid; done means the expected rollup behavior is understood and either documented or covered by a confirmed fix.

Written by the indexing model from the issue text.