The DMU does not "fix" issues automatically. It can be told to ignore them. For data with length issues, ignoring means truncating anything that does not fit. For data with invalid representation, it means the conversion may produce garbage for selected characters or for whole values, depending on what the problematic content is. The approach to ignore certain issues is not unreasonable seeing that you cannot make columns longer in eBS or change their data type. However, blindly ignoring all issues is not an option. You must carefully analyze the data and make sure that:
1. No data is binary stored in VARCHAR2. Such data would become totally illegible if converted without cleansing.
2. Data that is truncated or corrupted is legally irrelevant. Many financial records require a retention period of a few years. Truncating or corrupting important information in such records, e.g. contractor names, may be illegal.
For data that is important, you should use manual editing through the front end to fix the issues, for example by abbreviating the content in a meaningful way.