With an OCUM poller the site is just used for the meta metrics of the poller (how long the API took to respond, how many metrics, etc). Capacity metrics for the monitored clusters are submitted using their own site found in their own poller config. In this way the data is parallel in the Graphite metrics hierarchy yet seperated because the data has a different update frequency and retention.
I answered a few other questions about OCUM entries in this communities post, so maybe take a look there for a more detailed discussion including some examples.
Cheers, Chris Madden
Storage Architect, NetApp EMEA (and author of Harvest)