Reasons why DynamoDB TTL is deleting items 48 hours early

Question

I was looking to set up TTL for my DynamoDB table and set it to be the "TTL" key. Used python3 + boto3 to do a batch write with the TTL field set to str(int(time.time()) + 172800). I also tried int(time.time()) + 172800.

In either case, the epoch time stamp in the TTL column was 48 hours in the future. When using the second version, hovering over the value in the DynamoDB table showed a pop up with the date and time of the time stamp and I confirmed it was 2 days in the future.

However, when I came back ~5 minutes later and refreshed the table, all of the entries were gone.

I repeated the process and actively refreshed to keep an eye on the values and they were all getting deleted gradually with later time stamps being deleted last. I checked the cloudwatch logs and they showed my scans and writes in the correct GMT time.

I'm just wondering what might cause this to happen. For reference the table's creation date was July 27, 2019; is it possible that the clock for the db is off?