I don't believe FDM will allow you to have a blank value as a source value for a mapping. Depending on your version of ERPI you will have to handle this at import. For 18.104.22.168 in the import format use an import script to check for the existence of blank values and replace with an appropriate default e.g. None or Blank. For 11.1.2 use the same technique but you will have to put the check code in the importation event scripts as import formats are not used in this version.
We have had a workaround solution since moving to V11. This was not a problem in System 9.
The issue with blanks not defaulting to [None] is that some values escape the test. Such as forcing default values using "Like (* = [None])". If a new code is in your data that is not in your map, the Validation process foiled since [None] is valid and all blanks are now [None]. Only then do you either have an intersection violation that shows up during Export or you get bad data in HFM.
I don't like this new "feature" in S11 FDM. We will be submitting a request to have the default functionality added back. Perhaps as an option?
Just for your awareness 892775, we had used a similar approach when we did an upgrade of FDM, but we found that the conditional map scripts had a significant impact in the performance using this. We went down the path of putting a datapump script on the import format to make null fields actually return a value of "Blank" and then used explicit mapping to map Blank to [none] and it had a dramatic impact on our performance.
Just my 2 cents