If I deinterlace the movie (before AR-cropping and compressing) the result is worse than without deinterlacing.
Sure, but if you have already visually confirmed that video is not interlaced, what is the point to try it nevertheless?
The point is ...
as I stated before in this thread, that the movie is marked as interlaced.
And so "someone" will pay attention to that (Player, decoder, ...)
and deinterlace it (Actually it looks like MPC-HC produces a worse image
playing the original (!) stream than the non-deinterlaced movie produced by Avidemux).
And we're talking about a standard commercial DVD (

).
"It looks like" - I'm not sure.
And I find it quite time observing to manually find out for every movie
one handles if interlace is set correctly.
That's the reason why I asked about motion vectors etc.
But maybe this is like asking for a navigation assistant ...
in a racing car?
Rather electric windows in a bicycle. Look, Avidemux is what you make out of it. The source code is open, quality contributions are welcome. Make your hands dirty, reshape things.
... Thankyou for your invitation.

I hope you don't take it as a pretext, if I tell you,
that I take Avidemux as a tool to "make my hands dirty" on my things.
It's not my thing to work on the tool itself.
My contribution is to give a feedback/ hopefully unobtrusive hints
of a newbie.
So I disagree with your metaphor.
A bicycle without electric windows does no harm.
In contrast to a racing car without knowing how to deal with it.
It might well be possible that the majority of the Avidemux
produces films are bad in quality because most of the users
don't know about the pitfalls, two of them we are discussing
(DAR, interlacing).