Well, if you want answers from NVIDIA, this might not be the right place to ask - I cannot remember any employee posting to the OpenCL boards for some months. Besides, such statements are most likely reserved for the department of marketing, deception and confusion - not the technical staff.
Porting for sure is what NVIDIA would like to happen, but would bound your code to their hardware (or force you to maintain two accelerator platforms, C for CUDA and OpenCL).
Which features of OpenCL 1.1 does your code depend on? Some of the functionality can also be mimicked by additional user code allowing you to move back to OpenCL 1.0 - but of course that is not generally feasible.
PS: Next week will be the first anniversary of OpenCL 1.1 and NVIDIAs conformance candidate driver! Anyone donating a digital cake to celebrate?