If the field type of the value column in the data table passed to the parser is numeric, the parser converts the column numbers to text using the user's culture. The simplest way to avoid this is to set the field type of the value column to String. They system will pass the value as-is, keeping the source formatting.
A more roundabout approach is to add an xfbr rule to the Thousand Indicator and Decimal Indicator fields of the Data Source so they are updated dynamically based in the user's culture. The document Henning references contains an example should you need one, but it should not be necessary.