Mapping a partitioned type or its partition does not give the same result, data can be lost when partition has release character as shown in the code sample attached.Actually mapping the partitioned type does not mean "map the partition that was identify at parsing".
Explanation of the support is the following:
The partitioned type has different properties from those of the partitions. For example, Field has no associated syntax, as does its partition "NonQualified"; its partition "Qualified" has an initiator, terminator, and release character (all quotation marks). When we find an instance of one of the partitions, we say it's an instance of the partitioned type, without looking at the properties. If the content of the partition is at offset X for a length of Y, we say that the content of the partitioned type is also at offset X flr a length of Y. But in the case of the partition, if the data contains release characters, we remove the release characters and build a copy of the released data in a "Released Data Header" (RDH), and we set a flag in the work file entry for that instance to indicate that it contains release characters, so we'll get the data from the RDH when it's referenced. But we don't do that for the partitioned type, so you get the "raw data" (for the wrong length) when you reference the partitioned type.
The current behavior is error prone. The present request is to improve mapping code by considering that mapping the partitioned type would mean "map the partition that was identified at parsing".