While everyone correctly suggested there is minutiae of details happening in real-life, varying angle, slight offsets from geometry standpoint, dust/grime/etc.. from material standpoint,
I too have wondered that while I agree math is mostly always correct (though there are often exceptions, just remember the darkerning material edges), I wonder how complete it is, and if it would
be able to converge to absolute reality, if the user input was inhumanly perfect.
Even such simple stuff as GI caustic dominate our interior lighting in real-life so strongly that CGI will always look wrong, no matter how perfectly rendered without it. And that is just single factor, that
are many others such phenomena. One other is tailored BRDF since some materials behave too uniquely to light to be able to be correctly expressed by common model.
So, I think some user here, are maybe too quick to fault the user. Yes, some are too quick to fault the renderer too, but this is renderer's forum we should all strive for the rendering engine to be perfect alongside improving ourselves.
I've kep adding tons of detail to my wall modeling, and lot of complexity to my wall shaders at time, and I have never achieved even remotely similar results to walls that I am staring into everyday. It's not even close. Maybe to some,
but I've been looking this sort of detail for years and I can still see massive difference between reality and even the best available public pathtracers (only talking about shadow expression, not scene look). So far it was only the utterly slow and unflexible Maxwell that got decently close.