Ok, here it is. I have a rule:
If [all] of the following conditions are met
* Extension is pdf
* [Contents] [contain match] [bankAccount.findBank]
* [Contents] [contain match] [bankAccount.findAccountNo]
* [Contents] [contain match] [bankAccount.findBankStatement]
* [Contents] [contain match] [bankAccount.matchDatePre][date]
And I have an embedded table bankAccount that looks like this:
findBank,findAccountNo,findBankStatement,matchDatePre,rowId(just-for-reference)
Bank A,000001,Kontoauszug, Seite ,1
Bank A,000001,Mitteilung, Seite ,2
Bank A,000002,Kontoauszug, Seite ,3
Bank A,000002,Mitteilung, Seite ,4
Bank B,000011,Kontoauszug, Seite ,5
Bank B,000011,Mitteilung, Seite ,6
Bank B,000011,Entgelte, Seite ,7
...
Now some files don't get processed, because Hazel tries to match agains previous rows, that are very similar. For instance it tries to match rowId 3 where the proper document match is rowId 4. I can see that it tries to validate agains the wrong row, because of the "Custom Attributes" section, when I click on the Rule match indicator (green or red). It's the wrong row number.
When I move the row up, it matches properly. But then the other file will not be sorted.
That's also why I meant it would be helpful to debug agains a specific row of a table (UX problem). The red indicator does not help, because it's for the wrong row.