Thanks for reporting this. I investigated and the issue is that fapemig.br is a JavaScript single-page application (built on Nuxt.js/Vue). When NewsBlur fetches the page, the server returns an empty content grid with no actual card data in the HTML. All the chamadas are loaded dynamically via JavaScript from a backend API after the page loads in your browser.
Since Web Feeds work by analyzing the server-rendered HTML to find repeating patterns, there’s nothing for it to find on this page. The grid is literally empty in the source HTML.
The good news is that the backend API is actually a WordPress site that has its own RSS feed. You can subscribe to https://api.site.fapemig.br/feed/ directly as a regular feed, though it may not have the exact same filtered content as the chamadas page.
I’m looking into adding JavaScript rendering support for SPA sites like this in a future update.