Why diagrams matter more than they look like they do
Many claim limitations describe structure: a flow path, an assembly, an arrangement of components, a topology. The strongest evidence for a structural limitation is usually not in the prose of a product manual. It is in a labeled exploded view, a system schematic, or a cross-section diagram that shows the structure directly.
A claim chart that processes only the text of a product document misses that evidence and reads weaker than it should. A chart that picks evidence indiscriminately from diagrams creates a different problem: vague captions, unlabeled arrows, and OCR artifacts that cannot be verified.
Figures get the same verbatim discipline as text
PatentFlare reads figures alongside text, with the same standard of verifiability. Where evidence comes from a diagram, schematic, or labeled illustration, the chart cites the figure and page so the reader can confirm it at the source.
What the system does not infer from a figure
A diagram showing component A connected to component B is not extracted as evidence that A and B together implement a third structure or function unless that is what the labels say. The system errs toward extracting only what is explicitly labeled or captioned, on the theory that an inferred relationship should be the work of the attorney reading the chart, not of the system producing it.
This is the same principle that drives the audit pass on every disclosed finding. Evidence that requires an inferential step between the diagram and the claim limitation is downgraded from disclosed to suggested.