Help to understand clustering results

[Danilo Nunes]

I'm using the CluStream algorithm with the following parameters:

k = 5; micro-clusters number = 100

The stream analysed is the electricity dataset, it contains 2 class

And the "decay horizon" for the stream = 1000

I was studying the source code of "purity" measure, in its calculus it's used a "MembershipMatrix" class that shows where each data was clustered, using the probability inclusion of the clusters.

Also, in this calculus has a extra clusters for data that was NOT included in any of the real k clusters 

In the picture above, the highlighted in blue represents the extra cluster
It contains the most data of the stream window, and this happens for the entire execution

Is this a bad result?
If the answer is "yes", what should I do to get a better result?
I'm very confused to understading this

Thank you so so much

[Danilo Nunes]

Edit: Calculus = Calculation
Sorry for my poor english skills