Athena scans all data when querying for struct field

Hi,
Let's say I have an AWS Glue table with the following schema:

CREATE EXTERNAL TABLE `simulated_data`(
  `stream_src` string,
  `raw_data` string, 
  `my_struct` struct<date:string,time:string,value:float>)
ROW FORMAT SERDE 
  'org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe' 
STORED AS INPUTFORMAT 
  'org.apache.hadoop.hive.ql.io.parquet.MapredParquetInputFormat' 
OUTPUTFORMAT 
  'org.apache.hadoop.hive.ql.io.parquet.MapredParquetOutputFormat'
LOCATION
  's3://my-bucket/'
TBLPROPERTIES (
  'classification'='parquet', 
  'parquet.compress'='SNAPPY', 
  'transient_lastDdlTime'='1565760085')

The data is stored in S3 in multiple Apache Parquet files. The raw_data field contains JSON string with the same data (and structure) as my_struct struct. Example of data row is below:

{
    "stream_src": "src1",
    "raw_data": "{\"date\":\"\",\"value\":18.7,\"time\":\"5:30\"}",
    "my_struct": {
        "date": "",
        "time": "5:30",
        "value": 18.7
    }
}

I have a problem with querying the structure - Athena scans all data stored in the S3 bucket. The query that I use is:
Run time: 2.36 seconds, Data scanned: 1.88 MB (same size as select * from simulated_data )

SELECT my_struct.time
FROM simulated_data
WHERE stream_src='src1'

Surprisingly, the problem does not exist with JSON_extract
Run time: 2.02 seconds, Data scanned: 1.12 MB

SELECT json_extract(raw_data, '$.time')
FROM simulated_data
WHERE stream_src='src1'

What is the reason for that? Why reading field of structure leads to scanning the whole data set? Answering those questions probably will give me a light on why is JSON_extract faster than selecting data from the struct.