Two failures, one habit
Treat a blank as zero and a program looks cheaper than it is. Treat a blank as the average of the others and a program looks more expensive than the evidence supports. Both produce a number that no provider ever published.
The guide does neither. A field with no source stays null, renders as not established, and scores nothing in the component it belongs to.
The visible cost of that choice
Several programs in this dataset score lower than their price alone would suggest, because one fee is unpriced. That is the intended outcome: a comparison should show where the evidence stops.