I'll give you the same advice I give to everyone who says they have an issue with focus accuracy.
You have to eliminate the variables -- the reasons why focus might not be sharp which could be attributed to use error -- so that it's just the camera focus system being tested.
For that, you'll need to use a focus test chart and a tripod. I use a commercially produced chart called the "LensCal" (by DataColor) but there's also the "LensAlign" tool. There's a software calibration tool called "Reikan FoCal" which captures the shots and does the analysis for you. And lastly, there's Jeffrey Friedl's free focus test chart (downloadable from the internet here): http://regex.info/blog/photo-tech/focus-chart
For all of these, your camera will NEED to be on a tripod. You cannot trust a hand-held test because your body may move slightly forward or backward after locking focus which would invalidate the test.
You are also shooting a focus-test chart instead of a normal subject. There are a lot of reasons why you cannot be sure exactly what point the camera used to lock focus when using ordinary subjects. Since a focus test chart is designed to get the camera to focus at one particular point, you know what camera used for focus. The test charts also offer a way to determine (with repeatability) if the camera and lens are actually front focusing or rear-focusing (consistently vs. randomly) and by how much.
If the camera & lens are consistently missing focus, but nearly always in the same direction and by the same amont, then that's a simple focus calibration adjustment (that's why your camera has an Auto-Focus Micro-Adjustment feature). If, on the other hand, the camera is missing focusing randomly, then it usually means you've got some sloppy gear backlash in the lens and the lens needs service.
When I run focus tests, I deliberately de-focus the lens and re-shoot the target many times to force the camera to move the focus motor each time -- to bring the target to focus. For the first half-dozen shots, I run the focus in so that the camera has to run focus outward to lock focus. For the next half-dozen shots, I run the focus out toward infinity so that the camera has to run the focus in to lock focus. I do this because it's possible to see gear backlash in one direction more than another (e.g. it's possible that in one direction the camera might usually get accurate focus, but in the other direction it might usually miss focus.)
Again, you have to use a stationary high-contrast focus test chart. You have to use a tripod. You have to give the camera plenty of light. You have to eliminate every possible variable for why the camera might miss focus due to human error, target contrast, inadequate lighting, etc. etc. so that in the end, the only reason why the camera & lens could possibly miss focus would be a problem in the focus system itself (e.g. something you can correct via AFMA adjustment or the lens needs service.)
Tim Campbell
5D III, 5D IV, 60Da
