I had a source which had a 3-2 pattern, and I thought the combing artifacts were quite sharp (i.e., I really thought I had a perfectly telecined source)....however, Telecide failed miserably, and so I decided the next best thing would be to deinterlace.

I chose SmartDeinterlace, new version. I found that if I didn't use "blend instead of interpolate in motion areas", that I ended up with "banding" in the output video. And I also think that "frame-only differencing" gave the best results...I forget if I got "banding" also when using any kind of field-differencing. I just used the default 15 threshhold, and the default "compare color channels".

Can someone tell me why Telecide failed, and if my findings make sense (and why), and if there are any other SmartDeinterlace parameters I should investigate for a clip like this.