There are a lot of ways to think about zones as to how they describe field variability. I personally like to use 5 zones as a standard all the time. When I start sampling a field that is new to me, I might even sample all 5 zones. However, depending on the field, I will usually ending reducing the number of zones I sample, but I keep my map the same. ie- my map still have 5 zones, but I might only be sampling 3 (in the RRV, I actually have a couple of 5 zone fields that are now sampled with only 2 zones and it works well).
There are times when I will combine zones and sample them as one – ie-orange and red become one zone from a sampling perspective. However, I do this when I know the zones are behaving very similarly. I don’t like combining zones with very different soil characteristics (the extremes). I feel that these need to stand alone. However, if those extreme zones have been sampled routinely over the years and it seems like there is a very good understanding of the data and how that zone is behaving, it might not be necessary to sample that extreme zone every year. Simply use the previous years’ data to make decisions (which should work well except for soil mobile nutrients like N and S).  This is a good thought process especially when dealing with zones that are habitually low productivity and putting more time and effort into those areas probably isn’t worth it due to the low yield potential. 
In the scenario that you describe above, I agree with your assessment that blending zones 1 and 2 (if they are extremely different) can end up as a blended version and the results diluted.  Blending two zones with very different soil sampling results may produce diluted results that end up producing diluted data that really represents neither zone well.  In this scenario, I would rather keep the 5 zones and select which 3 zones are the best to sample and then when the results are in and then create the zone from this data.
So we have talked about either combining zones when sampling or sampling 3 of 5 zones but staying out of 2 zones when sampling.  Here are some options when creating the prescriptions in ADMS Merge Zone Results window. 

1. On Step 5 in Merge Zone Rezults:  Apply the correct zone ID to the zones that were sampled and select “NO Test” for the zones that were not tested.  This will force a rate on the zones with the assigned Zone ID and interpolate across the zones assigned with “No Test”.
2. If you have decided to combine colors (ie- zone 1 and 2 were sampled as 1 zone), You can assign those soil sample results to both of those zones by assigning the same sample result to both of the zones in step 5 of Merge Zones Results.

So there are definitely options in how to handle the data.  I think different geographies and fields might be different in how they want to handle it.  So I think find out what is right for you, the farm, and the field.  I have fields where I am doing this differently based on that field. 
As always, don’t hesitate to reach out if you have other questions!
Cheers!
Sarah