There are several issues related to Phase Detection Auto Focus performance. You first must determine what the source of the problem is. It could be caused by one of several factors, or a combination of some or all of them. If you also have the problem when using the Contrast Detection AF in Live View, then the problem is somewhere else.
Front/Back focusing. If the body lens combo consistently misses in one direction, this can be corrected using AF Fine Tune. The most accurate methods use flat targets parallel to the sensor plane. Tilted targets are great at demonstrating the concepts involved, but determining exactly what the focus sensor array is aimed at is more problematic that commonly thought. If your viewfinder says your focus point is over the "zero" point, but your focus array is actually focused on the "2", your adjustment will not be correct. Is the sharpest point always a little closer than the spot you wanted? That is probably being caused by front-focusing. If the sharpest point is always a little further than the point you wanted it is back-focusing.
Focus point location. The squares for each focus point in your viewfinder are only an approximation of the actual spot the focus array is pointed for each specific focus point. The corresponding points on the array are not physically arranged in the same layout as what you see in the viefinder. See here for a detailed look at the 5DII focus system. All multi-point AF systems behave this way to one degree or another. If most of your photos focus on a point in the same direction from what you are aiming at regardless of whether it is nearer or further away than your aiming point, this could be your issue.
Focus consistency. Phase detection AF developed over the years with the emphasis on speed over accuracy. The camera measured focus once, decided how much and which direction the focus needed to move, moved the lens, and then took the picture without any feedback after moving the focus mechanism. It was an "open loop" system. The technology has now matured to the point that manufacturers are also designing systems that do include feed back from lenses that tell the camera exactly how far they moved in response to the instruction the camera sent. Roger Cicala discusses this issue in his blog entry at LensRentals.com. To gain the advantages of a semi-closed loop system, both the lens and the body must have the capability. If there is no discernible pattern to the AF errors, it may just be the limits of the D7000 with that lens.
I would begin by doing a proper AF Fine Tune. With a zoom lens such as your 18-105, do it at the focal length you use the most. If you use a wide range of focal lengths, do it at the longest one you use frequently.
You can also do a "pattern test" like Andre did with his 7D. Understanding your AF systems characteristics will help you learn how to use it more effectively.
You might try a prime lens. Any zoom lens that covers a focal range of 18-105mm has design compromises that affect image quality and sharpness. You may be expecting too much from that lens.
Ken Rockwell says in his review of the Nikon 18-105mm VR:
The plastic-mount 18-105mm VR is a decent enough general-purpose lens for people who are in the price range of the D90 with which it is kitted, but for $400 ($300 in a kit with the D90), I'd rather buy something else.
The photos are nice and sharp most of the time, but if you're looking closely, the 18-105mm is Nikon's fuzziest lens in the corners at 18mm. Even the $100 18-55mm is better.
The DxO Mark Scores for this lens are pretty low as well.
DxO Mark scores compared to the Nikon 24-70mm f/2.8 and the Canon Ef 24-70mm f/2.8L II.
In response to some additional photo examples added to the original question and (since deleted) comments by the OP to this answer:
The additional examples added to the question bear out the point made in Roger Cicala's blog entry linked in the answer above: All PDAF systems have a range of focus accuracy from one shot to the next. The more accurate (and expensive) systems have a lower standard deviation, but they still vary a little from one shot to the next. Mount the camera securely on a tripod and use Live View to find the best performance your lens/camera combination is capable of. If the results using CDAF via Live view is about the same as PDAF, then you have found the limits of your lens.
If, on the other hand, the CDAF images are considerably sharper, the problem lies with your AF performance. Take 3 test shots of each position of the cans on the table and reset the focus to infinity between each shot. Then compare the three shots and see if the sharpest point of focus moves around or stays the same distance each time.
Thanks for your suggestions. I am not sure whats the intent of setting the focus to infinity? and you mean to do this in CDAF?
Use CDAF to see what the best your camera/lens is capable of. It will be slower but more accurate than PDAF. Then compare the results using CDAF to the results using PDAF. The focus variation comes into play in PDAF. Moving the focus to infinity between each shot gives an "honest" result when taking several shots with the same settings to establish the range of deviation.
