Some pages, instead of embedding structured data, only provide a reference to it for example with rel="meta" http://patterns.dataincubator.org/book/autodiscovery.html
Possibly also using rel="alternate"
`
`
or rel="describes"
<link rel="describes" href="https://data.alice.example/12312321441#this"/>
maybe sniffer could try following all such links using appropriate content negotiation?