An interesting journey and this is my take on things...
When you try and take the processing further than the data has to offer, time to stop.
Push it and you run the risk of adding detail that just isn't there.
No deconvolution in this version.
I find adding Ha & O3 to a galaxy quite difficult.
Remember that I am new to NB and have seen results from others that are truly stunning, so it can be done.
I tried several methods after lots of research and none cut the mustard.
Here is a technique that I worked out. It's not perfect but I think a better result.
After LRGB combination:
Take the Ha and remove the stars using Starnet (register to the LRGB beforehand).
Apply the histogram transformation to get rid of all noise and leave just the highlights (move the curve to the left).
Use Pixelmath to combine (Ha in this example):
R/K: $T+(Ha-med(Ha)) * scaling_factor (scaling_factor is a symbol and in my case = 0.5)
G: $T
B: $T
Drag the triangle onto the LRGB image.
Same goes for the O3 on the blue channel.
Time to open the beer fridge, one for you Mike & please continue to be honest
hires:
https://www.astrobin.com/full/o8iz5z/B/
...if you can blend the details in the previous version back in to the brighter core and cluster areas...I think it will look better 
...and THEN we can have that beer 
you'll get it, don't give up, this is what it's all about 
