On that regard, only the default settings have changed, the code and internal logic is the same.
Filmic midde grey setting is an exposure compensation. When you add 1 EV of exposure, you actually multiply the RGB values by 2^{+1\,EV} = 2, which is the same as opening your shutter twice as long (letting the raw black point/dark current aside).
In filmic, the first stage is to log-scale the dynamic range between 0 and 100% to scale the look curve, and we divide the RGB values by the grey value \log_2(x / grey) such that, for x = grey, \log_2(grey / grey) = \log_2(1) = 0\, EV.
By default, middle grey is set to 18.45 \% which is the default ICC value, but if you were to set it to 9.23 \%, it results in \log_2(2 × x / 0.1845) which, as you see, is equivalent to adding 1 EV of exposure (+1 EV in log space is equivalent to ×2 in linear space as per the logarithm properties).
That is because your cameras are probably using a dynamic range optimization in order to protect highlights, and “underexposing” the shot on purpose. Just add the corresponding +1.5 to 2.5 EV in exposure module, adjust filmic white and black bounds accordingly, and be done with it.
Filmic is tone mapping where a look is scaled to a dynamic range. Than can be SDR → SDR, HDR → SDR, HDR → HDR, or SDR->HDR, it doesn’t care. Filmic maps scene black, grey and white to display black, grey and white, no matter what the input and output values are.
Camera bias exposure is perfectly fine if you expect your camera lightmeter/auto-exposure to clip highlights if let alone.
It’s first and foremost the only color-preserving brightening method, that lets hue and saturation unchanged while affecting luminance.