One side of a comparison: static attributes, usable probe transcripts and run symptom fields.
probe:<id> and run:<field> read their own maps; an absent fact is null.
probe:<id>
run:<field>
null