Some digital cameras absorb IR light that is invisible to us. It's standard for digital cameras to put an IR filter in front of the CCD. Maybe your machine doesn't have such a filter for it's CCD or it isn't 100% effective.
I will have to ask out people if they have noticed that. I would think the effect is highly dependent on how the machine works.
Do you have an image based one or a rules based one?
I have noticed this on some of our large multi-circuit panels. It's such a slight rise in our false call rate that I haven't really explored it. I'm just guessing that it has something to do with the fiducials calculating the twist/shrink of the board while it is in a pliable state. Then, the inspection takes place after the board has had a chance to cool off and "snap back" to it's normal state. We use a Vi-technology 3K3, which is also an image-based machine.
I don't think there is anything releated to IR light in your false calls rate rise on Post-reflow.
Soldering Flux is most probably the reason.
Indeed, more or less soldering flux residue condenses on the boards according to the oven extraction system performances. And this flux, very hard to detect by any AOI, may also explain the false calls rise.
I'm also using a vitechnology machine and concerning the board warpage, their software includes a stretch correction option that adapts the inspection programs to the distance between fiducials. Plus, you can also use sub-panels fiducials to increase the stretch correction option precision. So warpage should not be a problem on this kind of equipment.