In some cases a single instance of blocking can ruin graphs for an entire month due to the enormous amount of wait time that can be accumulated in a short period of time. Below is an example of blocking that occurred on one of our servers for approximately 45 minutes.
Ignoring all of the queries that took a long time here (which is the only way to fix them currently) isn't ideal because those queries themselves aren't the problem and they could bear watching as they're likely very commonly run queries. Ideally I'd be able to filter out this ~45 minute period after I resolve the issue so that it doesn't make it more difficult to spot performance problems for a month following the incident.