Data vendors advertise coverage: hundreds of assets, dozens of venues. It reads as strictly more, and for spot prices it roughly is. For derived positioning it is not, because the computation stops being meaningful below a certain amount of real activity.
What thin books do to a map
A gamma profile computed over a chain with a handful of contracts outstanding produces enormous-looking concentrations at whatever strikes happen to have anything. The shape is dominated by noise, and it looks exactly like a shape dominated by positioning.
The same is true of funding on a venue with negligible open interest: the rate can be extreme and it describes the leverage demand of almost nobody.
The filter is the product
Deciding what is in the universe is a real piece of work and it is where honest vendors differ from ones optimising a coverage number. A threshold on open interest, a minimum number of active strikes, a requirement that the book has traded recently — each of these removes assets and improves everything left.
A vendor who publishes the threshold is telling you what the number means. One who advertises coverage without one is telling you how many rows they have.
What coverage is genuinely for
Breadth readings. Asking how much of the market is in a particular state needs a wide universe, because the answer is about the distribution rather than any one asset. A breadth figure over twenty coins is not breadth.
So the right arrangement is usually both: a wide universe for breadth and ranking, a filtered one for anything where you will act on a single asset.
What to ask
What is the inclusion rule, and can I see which assets were excluded and why. A vendor who can answer both has thought about it. A list of excluded assets with reasons is more informative than the included list.