With scripts, it's famously difficult to decide whether a page actually has "fully loaded":
On one extreme, a single page application may technically load very quickly (static assets fetched, DOMContentLoaded fired) but still be practically unusable until all asynchronous content has been fetched.
On the other extreme, a complex AMP page might look like it finished loading after a split second - but it's still fetching assets for the parts of the page outside the viewport.
Finally, caching: If caching is applied well, a page may be usable long before it completed loading.
So when they talk about "actual download speeds", I think it's important to know how they measured those speeds and what definition of "fully loaded" they used.
On one extreme, a single page application may technically load very quickly (static assets fetched, DOMContentLoaded fired) but still be practically unusable until all asynchronous content has been fetched.
On the other extreme, a complex AMP page might look like it finished loading after a split second - but it's still fetching assets for the parts of the page outside the viewport.
Finally, caching: If caching is applied well, a page may be usable long before it completed loading.
So when they talk about "actual download speeds", I think it's important to know how they measured those speeds and what definition of "fully loaded" they used.