Its difficult for us to change the source side column names as per the Fivetran normalization process.

If you handle space & hypen as underscore and already same column name with underscore exist, then please don't directly remove those columns by marking as duplicates.

Instead please keep it in destination table with some prefix like '_duplicate_column1_<Column_Name>' wherever it fails to suffice your normalization process.

So we won't lose any columns at destination. Please help us to get this change. 

Already I had a case with Fivetran #236070 and I got below response from one of your support engineer. Please refer below.

Fivetran's current naming conventions automatically convert spaces and special characters in column names to underscores, which can lead to columns being treated as duplicates if they differ only by these characters. 
 
This is a standard behavior to ensure compatibility across various data warehouses. 
 
Unfortunately, Fivetran does not currently support retaining spaces or using different special characters in column names directly in the destination.
 
To address this issue, you can consider the following options:
 Rename Columns at Source: Modify the column names in your source system to avoid conflicts after Fivetran's normalization process. This might involve using unique prefixes or suffixes to distinguish columns that would otherwise be normalized to the same name.
 
Submit a Feature Request: If the above solutions are not feasible, consider submitting a feature request to Fivetran's product team. Customer feedback is crucial for future updates, and your input can help prioritize the development of features that address this limitation. 
https://support.fivetran.com/hc/en-us/community/topics/360001909373-Feature-Requests
 
Unfortunately, without changes at the source there is no direct way to retain spaces or use different special characters in column names in the destination.