Comments (in languages, configuration formats etc.) are often used to encode metadata that only specific parsers can understand.
Perhaps the most well known example of this is Internet Explorer conditional comments where you can write markup targeting specific versions of internet explorer.
<p class="accent">
<!--[if IE]>
According to the conditional comment this is IE<br />
<![endif]-->
<!--[if IE 6]>
According to the conditional comment this is IE 6<br />
<![endif]-->
<!--[if IE 7]>
According to the conditional comment this is IE 7<br />
<![endif]-->
<!--[if IE 8]>
According to the conditional comment this is IE 8<br />
<![endif]-->
<!--[if IE 9]>
According to the conditional comment this is IE 9<br />
<![endif]-->
<!--[if gte IE 8]>
According to the conditional comment this is IE 8 or higher<br />
<![endif]-->
<!--[if lt IE 9]>
According to the conditional comment this is IE lower than 9<br />
<![endif]-->
<!--[if lte IE 7]>
According to the conditional comment this is IE lower or equal to 7<br />
<![endif]-->
<!--[if gt IE 6]>
According to the conditional comment this is IE greater than 6<br />
<![endif]-->
<!--[if !IE]>
According to the conditional comment this is not IE 5-9<br />
<!-- <![endif]-->
</p>
So not only the final parsed DOM that IE would construct would differ from what other browsers would construct, the parsed DOM could be entirely different among different versions of Internet Explorer.
This makes the job of people writing programs to parse HTML documents difficult because there is possibility that the program is throwing away valuable information as comments. Fortunately other browsers have not chosen to follow this trend.
JSON is intended primarily as an interchange format - that is simple, fast and decidedly NOT extendable. So it would make things complex if there were a mechanism to target specific implementations.
You can, however, always implement your custom extended JSON implementation (like MongoDB does for internal storage) and write your specialised parsers to handle that. Also you are more than free to build custom formats that are represented in JSON (like Falcor JSON graph which supports wrappers called atoms). But it is widely accepted in industry that if you are using something that claims to be a JSON parser and throw any document at it that claims to be valid JSON then the parsed outcome is exactly the same if neither of the claims are incorrect.