I have been reading the back and forth regarding the differences between Bitstreaming and LPCM and can offer a few thoughts. First off I will start with the assumption that there was an audible difference between the two for some of the tests that were related here.
Assuming the above one must consider what can cause it. For one thing, it is entirely possible that in the case of listening to an original LPCM track vs the DTHD or DTS-MA version the engineer did actually do something different with respect to how they were mixed. This would easily explain audible differences but only when comparing an actual LPCM track against the parallel Bitstream track.
In the case of a DTHD or DTS-MA track that is being decoded either on the card (or whatever source is being used) vs being decoded in an AVR then we have to look at how the decoding treats the metadata that is indeed in some of these streams. In the case of DTHD we know for a fact that it can include metadata describing how to handle things like dynamic range compression, dialog normalization and so on. I will post links at the end supporting these statements.
I cannot find any direct data supporting the notion of similar metadata for DTS-MA tracks however there is support for a sub-audio track that can have an effect on the output if used. It is possible for this to be a potential source of differences.
Finally, we come to the decoder itself. What we can be absolutely certain of is a case where we have the exact same chipset used exactly the same in both components - in this case I am talking more about SA players that have the same decoding chipset as the AVR/Pre-Pro it is attached to. This almost certainly cannot apply to the HDAV as I suspect it is a combination of hardware and software that is unique. The point here is that in this case we should have exactly the same sound as the decoded data is merely being transported over a wire from the SA that is a bit longer than the traces on the circuit board or inside the chip between the decoder and the DAC. I.e. LPCM is bit exact. Once the stream is decoded into LPCM it remains unchanged thereafter - as long as nothing else interferes (I am thinking of the Windows Audio Path). Any differences would have to arise in the decoding process.
One other potential difference in any case is that the AVR/Pre-Pro might treat bitstreamed data differently than LPCM and subject them to different DSP routines.
The upshot of all this is that while it is possible for a bitstreamed signal to sound identical to an LPCM signal, it is also possible that they are handled differently such that there is an audible difference. How often does this happen in real life? I personally am not sure as I have yet to hear any of this. I can state that a DTS-MA or DTHD encoded track that is decoded becomes an LPCM track. The question is whether or not any processing was done to the data. If so then this can account for the differences. If not, then the differences people are talking about become very hard to explain. Knowing how hard it is to set up a true blind test one would have to wonder at that point if there were subtle cues influencing the test. If it is not a double blind test with very sophisticated switching systems this is a very real possiblity.
Hopefully this helps people clear up this issue. We do not have enough data on hand to truly explain what people have related here, but one thing is clear. Depending upon your decoders you may or may not have a real difference. We actually can have a case where one system has an audible difference on certain material where a different system would sound identical. This is especially true in the case of DTHD as it is known to have metadata that can effect the decoding and different decoders can indeed handle this differently.
Finally, the best test of all would be to perform a digital comparison of the final output just prior to the DAC. This is really the best way to identify differences; if there are any then they would be easily identified. The catch here is that one would have to have quite a lab to accomplish this...
The DTS site has a couple of whitepapers on it that help but do not really answer the metadata question...
The Wiki page on DTHD explicitly states that there is metadata used for dialog normalization and dynamic range compression with perhaps other options too:
Hopefully this information allows us all to move on and accept that some might hear differences and others might not. As always YMMV.