Your citation pretty clearly demonstrates that drug-sniffing dogs DO work. And anybody who has ever had a dog sniff something out would find this to be an obvious fact.

The omitted figure is how large the sample size is. If the dog sniffs 100,000, and identifies 10 as having drug odor, but only 2 of the 10 are currently in the possession of drugs, that doesn't really indicate that the dog "has a failure rate of 75%" because it doesn't take into account any of the "negative identifications." We don't know how many of those other identifications were false negatives, but certainly I wouldn't expect 25% of the population to be carrying illicit drugs at any given time, and especially not in scenarios where there may be a checkpoint with a dog involved.

The dogs may be misused, but that doesn't make their ability a myth, and neither does this article claim that a dog can't sniff out the trace scents of a drug. It carefully skirts around this assertion, because dogs absolutely can sniff out drugs. They're just not infallible, and may also be used as pretext.