Looks like I have to register at every forum to get points across. Anyway:
Comparing the screenshots of NewRisingSun's algorithm to the MobyGames screenshot linked from that thread, it seems he has red artifacts where the MobyGames shot has blue, and vice-versa.
That's not a function of the algorithm, but of the ROM font used --- apparently the person who made the MobyGames shots used a graphics card that shifted the font by one pixel. The "O" character looks different, too. Also, my original algorithm reduced the artifacts to the level they occur on a TV screen, but that would require 16/32 bit rendering, and I'm told DosBox at the moment only supports 8-bit palettized rendering.
The text is readable I would say.
Those screenshots have "flagOptimizeText" disabled; you said you wanted to keep it enabled, only disable "flagFilterChroma" because of DosBox's restrictions.
Is it supposed to be that much darker?
Yes. The brightness is controlled by the lower four bits of port 0x3d9. KQ1 writes 27 to that register, so it's at half the brightness; other games (BC's Quest for Tires, Winnie the Pooh in the Hundred Acre Wood) write 3F to that register, so they're at full brightness.
I prefer the darker variant so I coded that in as default.
You're not supposed to code any value as a "default", but to use whatever the game writes to port 0x3d9.
although it will always be a lot darker then EGA. But yes. CGA 16 is darker. partly by design. partly by me (a tinybit).(4 % or so)
No, only if you implemented the algorithm incorrectly.
I'm not sure whether that CGA screenshot taken from the Oldskool PC site is shown at the proper intensity.
No, it isn't. It's supposed to be at half the intensity in "RGB mode".