I could see a circuit effectively giving increased frequency-dependent attenuation, so the noise is not higher, it's that the signal is lower so the signal-to-noise _ratio_ is lower. (Since it's in dB, this must always be a ratio, relative to some other value or to something fixed that is an understood reference level.) I'd need to look at the definition of QLN. I would think it doesn't fit this idea, you'd expect it to be relative to some fixed reference level, so it's effectively like an absolute noise level, rather than an SNR. I'm totally confused now.
A filter can let more of the noise through, of course. So the post-filter absolute noise figure goes up when the filtering is less effective, for example.