When AppFlow writes parquet to S3, can it maintain the source datatypes or not?

In the AppFlow UI, it appears to indicate that maintaining the source datatypes is possible via the option underneath parquet:

However, in documentation (https://docs.aws.amazon.com/appflow/latest/userguide/s3.html), it states:

If you choose Parquet as the format for your destination file in Amazon S3, the option to aggregate all records into one file per flow run will not be available. When choosing Parquet, Amazon AppFlow will write the output as string, and not declare the data types as defined by the source.

These two sources conflict with each other. The behavior I am seeing is what the documentation describes and all data is being written as string type. I am trying to determine if this is intended or a bug. If the latter, I can open a support ticket.