boto DynamoDB: How can I prevent overwriting items with batch_write?

I'm using boto's DynamoDB v2 and I'm writing items to a table in batch. However I'm unable to prevent DynamoDB from overwriting attributes of existing items. I'd rather have the process fail.

The table has the following schema:

from boto.dynamodb2.table import Table, HashKey, RangeKey

conn = get_connection()
t = Table.create(
        'intervals',
        schema=[
            HashKey('id'),
            RangeKey('start')
        ],
        connection=conn
    )

Say I insert one item:

item = {
    'id': '4920',
    'start': '20',
    'stop': '40'
}
t.put_item(data=item)

Now, when I insert new items with batch_write, I want to make sure DynamoDB will not overwrite the existing item. According to the documentation, this should be achieved with the overwrite parameter from the put_item method of the BatchTable class (which is the one that is used as context manager in the example below)

new_items = [{
    'id': '4920',
    'start': '20',
    'stop': '90'
}]

with t.batch_write() as batch:
    for i in new_items:
        batch.put_item(data=i, overwrite=False)

However, it doesn't. The stop attribute in my example gets a new value 90. So the previous value (40) is overwritten.

If I use the table's own put_item method, the overwrite parameter works. Setting it to True replaces the stop value while setting it to False results in a ConditionalCheckFailedException.

How can I get that exception when using batch_write?