The Input Features contain clusters of disconnected groups of features. There are more of these clusters than the Number of Groups you specified, and the Spatial Constraints parameters do not allow these clusters to be combined. Consequently, there are more groups going into the analysis than you've specified you want to come out of the analysis.
While you can increase the Number of Groups specified, realize that the Grouping Analysis will begin with the disconnected clusters already representing groups. For those groups, the spatial relationships alone are determining group membership (there is no analysis of feature attributes for those groups). The Spatial Constraints parameters that you specify determine which features can be grouped together and which features can't. If you are seeing a message about clusters of disconnected groups, you may want to relax your Spatial Constraints parameters. Selecting the K Nearest Neighbor Spatial Constraints and increasing the Number of Neighbors parameter until there are no disconnected groups will ensure that both spatial and data relationships are incorporated into the grouping algorithm. More information about how the Grouping Analysis tool works with disconnected clusters of features is described in the tool help.