That is not an accounting error. Cracking breaks large molecules into smaller
ones, and smaller molecules take up more room, so a refinery returns more
volume than it receives. The EIA reports it as a line item called processing
gain, and at the foot of the diagram you can see it: the red bar runs past
the black one.
Below is every major step between the two, drawn as descent. Crude at the top,
one ancestor. The first split is the only purely physical one in the whole
process; everything after it is chemistry. At the bottom, lines rejoin,
because the gasoline in your tank has at least four separate lineages.
How to read it
Red means measured. Every red figure is an EIA refinery yield.
Nothing inferred or drawn for structure is ever red.
Width means volume, and only at the two ends. No line in the
middle carries a quantity, because no intermediate quantity is published.
Drawing one would be a guess wearing the costume of a measurement.
Dashed red is hydrogen. It runs backwards up the diagram: the
reformer makes the hydrogen that cleans its own feed. Refining is not a
tree, and that loop is why.
What is not here
The visbreaker
US visbreaking capacity is 16,000 barrels a day against 2,793,645 for
coking. These yields are American, so a unit at 0.08% of the barrel is left
out rather than drawn at a scale it does not have. In Europe and Asia it
would belong.
Miscellaneous products, 0.5%
EIA's residual category. No single lineage produces it, so it is named
rather than given a parent it does not have.
Volumes through any individual unit
Not published. The yields at the ends are real; the path between them is
documented derivation, not measured flow.