Yes, C4D does not correctly render 32bit Alpha and messes up when embedding it in 32bit formats. Render the Alpha as a separate sequence and use matting techniques. NaN is for "not a number", meaning an invalid value format.
Mylenium certainly knows Cinema 4D better than I do.
Lutz, are you saying that this is a bug in Cinema 4D?
Dan, have you seen this issue with mattes generated from other software, or just Cinema 4D?
Thanks for the flowers, but I've only recently started to use that sucker again after years of staying away as best as I could (though even then I helped out some friends occasionally) I'm reasonably sure it is an issue in C4D and they are aware of it. I seem to recall that it was "fixed" in one version, but then broke again when they started reworking the render engine in recent versions. I think they're still treating Alpha as 16bit and that seems to conflict with some settings for antialiasing and sampling. Also using some of their render effects like Vector Motion Blur has been known to show this behavior. It's not occurring per se, but seems to be tied to specific combinations of things. Might wanna advise your peopel to chit-chat at NAB about it (and snatch a test license perhaps for investigating?).
I think that the best thing for us to do, then, is to put this in our bug database and then use our usual process for handing off bugs to other software companies. In other words, I'll take it from here.
Thanks Mylenium and Todd for confirming what I already suspected - C4D is buggy as hell, especially in versions <R12.
Can you clarify if you've discovered this behaviour as it applies to 32bit object buffers? There's no "alpha" in the scene; the footage is luma matted with the greyscale object buffer from C4D.