Well it seems to me that's what will always happen if you link on Date. Rather, if your second table is for forecasting, you should only be linking on Item. Rename Date from Forecast to [Forecast Date] or something. It's not possible to exclude a table from a loop-and-reduce; even if it were possible though, it would not be a very elegant solution. You would typically want to avoid synthetic key situations like the one you're introducing here. Once the data model is revised a little your problems should go away.
thanks for your response. I think you are right with the recommendation. Of course my data model is not that easy as described in the example, but I will try to merge the "two" tables and ignore the date.
I don't know of any way to set patterns on only some bars or to adjust the bars width. Maybe you can play with the new Gandtt chart available in v10, it might work better for you. Meanwhile, I'm attaching the solution that I mentioned before that I made from some fake data--I think it's a very decent view and you can easily tell from the legend what the red bars represent.