A few honest limits, because a table this confident-looking should carry them visibly.
The mercury figures describe species, not cans. FDA's numbers are means from sampling conducted between 1990 and 2012, and brand-to-brand variation within a species is real and documented. The table tells you what a fish tends to carry, not what is in the specific tin in your hand.
Some of those figures are also old. Every mackerel number in the FDA table traces to a 1978 National Marine Fisheries Service survey, or to a 2000 Gulf of Mexico report in the case of king mackerel, rather than to FDA's own later sampling — and the sample sizes are modest, at 80 fish for Atlantic mackerel and 30 for Pacific chub. They are the best federal numbers available and they are decades old, and both of those things are true at once.
Sample counts vary sharply elsewhere too. The canned salmon mercury figure rests on just 19 samples, against 545 for canned light tuna and 451 for albacore. That does not make salmon's low reading wrong, but it is a thinner foundation than the tuna figures stand on, and 0.014 parts per million should be read as an estimate rather than a precise value.
There is one seam in the sourcing worth flagging as well. FDA's table lists "sardine" generically, without specifying canned, while the nutrition figures come from a canned record — so the sardine row joins two datasets that do not match perfectly. Finally, prices move. The August 2026 snapshot here is reproducible because the retailer, the products and the date are all named, but the per-dollar figures will drift, and the ordering is the durable part.