Sometimes the triggered backup is missing table directories

We take backup of our Scylla cluster on a regular basis and push it to a S3 bucket.
But recently I has started to notice that one of the backups is having only 2 snapshot table directories rather than having 3.
Tables in the keyspace directory of the cluster
```
[root@scylla-lst-visual-classification-4-1a] ls /data/scylla/lst_image_classification_details/
imageclassificationdetails-68947a00287011e98990000000000005 messageoriginaltext-6eeced60287011e98990000000000005 parenttochildmessagemap-73655620287011e98990000000000005
[root@scylla-lst-visual-classification-4-1a]

```
Directories uploaded to S3 by the tool on two different days (backup on 20-02-2019 is missing table named `parenttochildmessagemap` )

```
[root@build-master-2-1b] aws s3 ls prod-scylla-backup/lst-visual-classification-20190220/lst_image_classification_details/
PRE imageclassificationdetails-68947a00287011e98990000000000005/
PRE messageoriginaltext-6eeced60287011e98990000000000005/
[root@build-master-2-1b] aws s3 ls prod-scylla-backup/lst-visual-classification-20190219/lst_image_classification_details/
PRE imageclassificationdetails-68947a00287011e98990000000000005/
PRE messageoriginaltext-6eeced60287011e98990000000000005/
PRE parenttochildmessagemap-73655620287011e98990000000000005/
[root@build-master-2-1b]
```
Following is the command which we use for triggereing backup
```
scyllabackup take -c /etc/scyllabackup.yml --prefix=lst-visual-classification-20190220
```
tried triggering the `take` command with `-l DEBUG` but couldnt find any issues .
Can someone please suggest , what might be triggering this anomolous behaviour ?