Did this thing ever work? Did anyone verify that the queries were reasonable and that the results were correct? If it did work, what changes (environment etc.) preceded the failure?
You've got several issues. The first being this error.
The second is that just from what you have reported the model is a dog's breakfast. It's possible that the bits which you do not mention are involved somehow too.
==> Are we delaing (sic) with what we call a LOOP here in your opinion ?
Yes probably a loop. Not a good thing but that should not need to be restated.
Why does this dimension to dimension relationship exist? You don't say, which precludes any analysis.
==> I don't know, i (sic) started using that cognos package a couple of days only.
It would be helpful for you to know.
Since you mention that Sales also has facts (OrderTotalAmnt), it is probable that it is a fact table as well. You need to be able to identify the dimensions and facts. You don't mention the cardinalities of the relationships, which is probably something which you should.
What are the properties of the query item which causes the problem?
==> [zip Code]
Usage is an important property.
==> Exactly, there is a FILTER using a field from [DIM TIME] and the filter is as follows:
[QuerySubject].[Dim Time].[QuarterYear] = 'Q3-20'
but the error still displays even when getting rid of this filter IF we add the specified problematic CHAR field in the report indicated before.
The same error or another one?
What exactly do you mean by 'factual' when you talk about things like SalesOrder# and Invoice# ?
Those are numeric columns but are not facts; they are identifiers. If their usage have been set to fact (aka measure) then the query engine will treat them differently than if they had the correct usage. It is possible that whoever did the modelling of this did not check the usage after metadata import. This is one of the proven practices.