To me the first set of images looks like just not being indefinitely sharp - which is typical for real-live images: The optics in front of the camera sensor isn't perfect, even if you pay thousands of Dollars/Euros for it. The autofocus of your camera is optimized on something between speed and accuracy. And even if it happens to tell the optics to keep exactly this point in focus . the rest of the scene, being 3D, won''t be. Then the image is compressed so many of them fill on real-live disks - and only every roughly every fourth pixel gives you a measurement for the "red" value, every fourth pixel one for the blue one and every fourth for the green one - even without the pixels on the sensor being so small that it is virtually impossible for a photon to make sure to in the end always hitting the correct one That phenomenon is then counter-acted by the software that interprets data from the sensor.
The result is: You get pictures that are sharp enough for most real-world applications. But if you zoom in and look at the details you see that no mechanism human beings create is perfect.
The last image you sent looks like you moved the camera while it was shot. The darker it is (and thus: the longer the camera has to collect light in order to get enough photons per camera pixel that it can offer a good signal-to-noise ratio) and the more your optical zoom zooms enlarges the picture (and this reduces the area to collect light from whilst enlarging any effect from moving the camera) the harder it is to get a steady image.