It is highly recommended to have version control on output (RDF) files and perform conversion incrementally. The following workflow is proven to be a good practice:
Typically, a change in converters is small in terms of source code, but it makes changes all over the data, in various places. It is very difficult to oversee these changes, especially in datasets with millions records.
It makes sense to work in small steps, and do one thing at a time. Then the changes are easier to check with diff on the output RDF files.