Hi Aerospace Mfg
You have certainly made some progress.
Not an easy one. The reduction in oxide clusters probably has little effect on hardness overall, but the increase in the finer (normal) oxides will increase hardness slightly. To reduce unmelts further, you need to get more heat into the particles. This unfortunately may result in an upward trend in hardness.
My thinking now is directed to powder injection, obviously if this is not set to optimum conditions, then some of the powder may not be positioned properly in the plasma stream. The more massive fraction of the powder will tend to penetrate the plasma further than the lighter fraction (classification effect). I would try playing with different carrier gas flows to see if you get any improvement. Also check powder port condition and alignment. I know 444 can have a nasty tendency to foul powder ports and nozzle faces particularly on long runs. If you notice any spits or pimples on the coatings, then this is where they come from.
A little puzzled why a maximum hardness is specified. Also, we need to be careful how we define what is an umelted particle. May worth discussing with your lab, just to make sure you are not rejecting coatings unnecessarily.
What were the original and modified spray parameters? Have you tried the standard Metco parameters for the 3MB/7MB (not keen on high spray rate 35lb/hr? 9MB parameters as far as I can remember). That's assuming you are using these guns