For some feeds, the LumenWorks.Framework.IO Csv does a better job than the home-grown ParameterizedDelimiterParser. But...it turns out there's a bug in the Microsoft Xml deserialization code when trying to specify \t in the plan. So instead we just derive from the CsvParser here, and hardcode the \t.