I would suggest an extra enumeration value like ‘unknown value’ and use a microflow on import to parse the value in the csv to the correct enum value. If it's not one of the known values of the enum and not empty, you could set the ‘unknown value’. Perhaps map the unknown value to a helper entity or attribute to still have the data in your model.
But if your expecting more of these ‘unknown’ values, I would refactor my enum to a entity of its own. Then you can create values in runtime as needed.