Pages with different URLs and no rel=canonical tags are treated as identical

### Steps to reproduce
1. Open an article on getpocket.com
2. Annotate it
3. Navigate to another article and annotate it.

### Expected behaviour

* annotation extension (more likely the server, see below) should detect the URL & title correctly
* as a result, annotations are split across multiple pages
* as a result, no annotations get orphaned

### Actual behaviour

On step 3, the URL in the navbar & the title tag are changed by the Pocket JS code. But the problem persists even if you reload the page manually.

I couldn't intercept the AJAX requests of the extension via the Network tab, might try with Wireshark later.

Here is the interesting part: the extension seems to send the right URL to the server but it groups it under a single URL & title anyway:

I suspected `rel=alternate` or `rel=canonical` in the webpage source but didn't find any. Perhaps Pocket does some advanced DOM manipulation after loading, that I can't reverse engineer, because there was nothing suspicious in the HTTP headers either.

### Browser/system information

macOS / Chrome / uBlock Origin