The 14 mOhm you mention is valid when Vgs >= 10V (in other words, the FET is fully turned on). The graph you show illustrates the FETs behaviour for Vgs in the range 2..8V, the range in which the FET is between fully off and fully on. In this operational mode the Isd is determined mostly by the Vgs (the FET behaves as a voltage-controlled constant current source). For this mode to work there must be a minimal Vds, otherwise that voltage would be the limiting factor for the Isd. Hence the Vds >= 10V requirement.
The input leakage current has various contributions. Here are some of them
1) the input protection circuitry (typically diodes).
2) parasitic surface conductance (very low) of the package.
3) leakage current, through the gate dielectric of the CMOS (This is very very negligible, unless dealing with very thin-oxides)
4) Other leakage sources, such as at the surface of the silicon die, between bonding wires, etc. These are all very low.
The largest source of leakage is 1 and, in some minor extent 2 (This depends on the surface contaminants). The other are by far very negligible. Cause 1 is also very sensitive to temperature.
I found that typically the producers are very conservative with their specs, and 10uA was some orders of magnitudes larger than what I characterized with a parameter anaylzer (B1500A), at least at 25°C.
Of course, a "large" input leakage can also be present on some input with pull-up, or on some logic with bus-hold function, but this is not the case of your IC.
The direction of the leakage current is not known a-priori. It depends on the external voltage applied to that pin.
Best Answer
This difference could be explained in a number of ways, though none definitive without knowing more about the device than the datasheet (necessarily) reveals. These may include differences between the actual structures of the underlying devices within the packaging (most likely), as well as differences in the estimations, testing practices, and / or acceptable margins of the companies or even specific engineers involved in their design and documentation. It is conceivable that typos in the actual datasheets themselves could be blamed. Nothing is certain until it can be proven that it works, and even then, things (manufacturing processes, etc) can change over time without you (or even the distributor, documentarian) necessarily being informed. Good testing procedures, performed on all devices prior to distribution (i.e. quality control) and wide operational margins for parts selection can help to insulate you from problems inherent in both mis-specification and changes in manufacturing over time, but don't be surprised if your (initial) design requires re-work somewhere along the line because of these things.