In the AUTOSAR System Template I see for some rows in tables have Note Section with Stereotypes: and Tags: but for not all rows of the table. So what is the difference between these two types of rows with respect to that Note comments?
In your example, the aggregation is stereotyped atpSplitable and atpVariant. The usage of atpSplitable implies that the aggregation can be split over several physical files and the partial models can be merged in the process of loading the model into an AUTOSAR tool.
The usage of atpVariation indicates that the aggregation undergoes a model transformation on the way to be generated into the AUTOSAR XML Schema and thereby acquires the infrastructure for defining multiple variants of the aggregation.
The stereotype atpVariation is in most cases combined with atpSplitable, but not the other way round. There are use cases for using partial models without variation, too.
The tag (atp.Splitkey) in the class table contains an instruction for the merging of the partial models in different files.
The meaning of these tags and stereotypes is documented in the AUTOSAR document "TPS Generic Structure Template".
Only a limited number of aggregations (or references) have a use case for being variant. All cases that do not support variation therefore lack the appearance of the stereotypes in the class table.