Actually I don't understand the attempt of the "preboost" within the source here - doesn't this 3dB adjustment affect all channels anyway? Then it is up to the listener to set the whole "gain" (and depending on the source then effective "volume").
This would only make sense for me if everyone at home would listen to a precalibrated system (including ideal room acoustics) - which I highly doubt - and expecting everything to play according to director's/mixing engineer's "vision". Otherwise, it is selected to personal taste anyway and easily changed at the AVR if necessary.
So I don't see why driving a source into clipping (and by that making it technically worse* as Turisu correctly pointed out) is even an option worth considering. Same is true by the way for the Cinema DTS decoding. Some LTE tracks for instance clip if the official gain recommendations are applied but the actual cleaner way would be to simply reduce the level of all other channels accordingly.
* I am aware that some clipping, especially in the case of the LFE will probably not be noticed but on the other hand, there is no good reason to provoke it even for one sample. PCM's performance at 16 bit already (properly dithered) is so good that even staying 6 or 12 dB below fullscale the whole time when in doubt and effectively reducing it to 14-15 bit, is far from being audibly apparent, not so speak of 24 bit where clipping is even more idiotic.