In order to get the class identification in HCP, regardless if you start at well level data or cell level data, you must have enough wells of positive and negative controls. Single cell level data is aggregated up to well level BEFORE the feature selection, HIT classification etc is complete.
The only way around this is to run your data as ‘non-plate object based’ data. Then you can use the control labels and it will apply those controls to every cell in those wells and it can be run that way. You should still still see the PCA, feature selection, HIT classification, SOM etc but you will not get the plate-based visuals because it is running as object data.