Isn't this a fundamental "risk" when using HDMI ARC as the source for a stereo amplifier, regardless of what amp that may be?
My perception is content sources differ in how they choose to "mix" a stereo output from a native 5.1 (or higher) original audio tracks. So you may get more or less center in your L/R stereo signal, and center is where dialog is often predominantly placed.
I've noticed this myself using an NAD D3045 connected via ARC from my LG TV. The other day I was watching a WNBA game on the Ion network, and the dialog from the announcers was very low, while the background noises from the court & stadium were unusually loud. I speculated that Ion was simply feeding L and R from a 5.1 native source to the L and R for a stereo downmix, rather than mixing center into both L and R for the downmix. When I reverted to my AVR and used true L/C/R signals, the dialog was fine.
I'd like to get rid of my AVR, but situations like this make me think twice.