PI AF attributes (used for analyses) lacks the ability to filter for conditional values. (e.g. the function PIAdvCalcFilVal). Often data can be represented poorly for calculations and there isnt an easy way to filter the outliers and use that value into PI AF analyses.
The infrastructure/functions for more complex data extractions/retrivals exist in PI Datalink but they aren't incorperated into PI AF. I think PI AF would benefit greatly from this functionality for attributes.