Dear all.
I received an update regarding my OSS note and it seems that the "expected" logic works like assumed, if you follow SAP's expectation 
:
I've checked the DTP and technically it works correctly.
In 7.3x/7.40/7.50 there is the possibility to have a choosen field
list for the start/end routine. When i logged on, the start routine
had more than 40 fields in the field list, but only 23 fields of the
source cube had assigned rules to them.
As the data is read in an aggregated form (just like in a query), the
number of records depends on the fields in the drilldown. As the
number of the fields in the field list of the start routine were
almost double of the number of fields which had rules assigned to
them, the system could not aggregate that much and therefore the
number of extracted records was bigger.
After i've adjusted the field list of the transformation, the DTP now extracts
the same amount of data you mention in your attachment even if there
is a start routine. The field list of the startroutine works exactly like a query.
For example if you have only 0CALYEAR in the drilldown, only one record is
displayed for each year.
If however you additionally add 0CALMONTH, there will be 12 rows for
each year. The same applies for the field list of the startroutine.
That means, "just creating" a startroutine is not sufficient. You need to adjust the list of used characteristics for the startroutine. If only the fields of the target (Infosource or Infoprovider) is chosen, then the DTP selection works as expected.
BR, Martin