There are a few ways to approach this, depending on the particular structure of the fields.
Some of those solution involve additional transpose steps to flatten out the field structure.
One we use to maintain structure for unknown fields is to join the output with a text input that has all of the possible fields with a single row of nulls. Then we throw out that row and you're left with your original data with all the expected fields (with null values). Then you can write formulas using the field that may or may not be present and wrap an IsNULL([Mysteryfield]) around them so they produce values either way.
There is also likely some way to do this with conditional execution, but that seems over complicated for the case.
Attaching a simple version of the text input/IsNull() method. If that doesn't help, we'll likely need a more specific picture of what you're trying to do (or possibly do) to zero in on which way to handle the issue.
