FiveTran replication when on Salesforce data, automatically transform source table with columns using internal FiveTran defined rule sets and create tables with columns at customer target with names transformed. 

This is causing the data loss on following situations on SalesForce as data source:

1) if multiple fields (columns) exist with different forms of names, such as Employee_Change_in_StatusAllowed__c and Employee_Change_in_Status_Allowed__c (this is a live example from our 'lead' SF table) FiveTran detects them as duplicates and pick one to replicate while ignore another. In our live example, FiveTran picks column with older data so we don't get the the newer data from another column. This is causing the data loss. Similar situation could also exist if source if case-sensitivity is enforced at the source and multiple columns with the same name but different cases in place that would also cause FiveTran's treatment as 'duplicates'.