Muxman's delay differs from the inputted one.

[Hyper Shinchan]

Hi, it's been ages since my last post, anyway I've been messing around with DVD reauthoring and I've come with some problems with the audio delay.
Basically PgcDemux gives me a certain audio delay and after having demuxed the tracks I remux them with MuxMan inputting the value that PgcDemux gave me (it's always a negative value in the DVDs that I'm authoring) .
Now I've noticed that if I open the new reauthored DVD with PgcDemux the delay is reduced compared to the one I've inputted (for example from -231 ms to -7 ms).
Is it normal? When remuxing the tracks I add some subtitles but I doubt that it can be related. Is it possible that PgcDemux gave me wrong delays and MuxMan "fixed" them?
I'm pretty new to DVD authoring and I'm quite hopeless about it, especially audio delay in DVDs has always been a mysterious topic for me.

EDIT:
I've tried remuxing the tracks with IfoEdit remux tool and it kept the original delay.

[bigotti5]

http://forum.doom9.org/showpost.php?p=819926&postcount=145

[Hyper Shinchan]

Thank you for the explanation bigotti5. But why do Muxman behave in this way? Is it because of requirements of the DVD specifications or player compatibility?

[manono]

You could always get rid of the delay entirely using DelayCut on the AC3 file. That's what I do.

[Hyper Shinchan]

Actually I'd prefer to keep the audio delay because I'm dealing with splitted discs but Ifoedit muxer seems quite poor of features compared to muxman (I'd especially need the wide/letterbox subtitle feature of muxman).

[manono]

Actually I'd prefer to keep the audio delay because I'm dealing with splitted discs...
I don't see what that has to do with anything. Using DelayCut you can both remove the delay and keep the audio the same length quite easily, if you think that by removing the delay it'll make the joined part go out of synch.

I think bigotti's link explained it anyway. You can only split on AC3 frames, each of which is 32ms long. Unless your delay is some positive or negative multiple of 32ms, then there will always be some some small delay left over. And as you know, -7 ms (your example) is way to little for anyone to notice since it's usually said one can't tell delay problems of less than 100ms.