georges
You have hit on one of the more interesting problems with Bing CV carbs.
The steps between needle positions are much too big in my opinion. I believe one needle position equates to 2 or 3 changes in needle jet size. So changing the needle jet will make a smaller change to the mixture than moving the needle.
Where it gets very difficult is comparing different needle positions with different needle jets and putting them in order of mixture strength. For example: A 2.66 jet with needle position 3 is obviously richer than a 2.64 jet with the same needle position but how do you compare a 2.66 position 3 with a 2.64 position 4 ?
I have seen it done mathematically on a spreadsheet. It involves a lot of work measuring the profile of the needle at small increments along the length of the tapered section. Then you have to calculate the area of the needle and subtract that from the area of the needle jet for each increment. Do that for each combination of needle and needle jet and you can make real comparisons. As far as I know it hasn't been done for a BMW Bing CV but if you are interested I can dig out a document that describes it being done on a Bing carb in another application.
There is another way you can achieve position 3.5 which I have done for my carbs (except I was looking to run at 2.5) and that's by shimming the needle jet height There is the issue of this also changing the jet height relative to the fuel level in the bowl but it does work. You can follow a description of it on a previous thread here
http://www.bmwr65.org/cgi-bin/yabb2/YaBB.pl?num=1254577036/0#0