1. Shoot in raw - saving to JPEG has already discarded image information, thus limiting post-processing flexibility.
2. The extra DR can not be used directly with current output devices (displays & printers) as their DR is much smaller. Where the extra DR comes in handy is when you boost the dark areas in post processing e.g. a group where some of the people in a brightly-lit image are in shadow, backlight or otherwise poorly lit. In these circumstances the dark areas are less likely to end up looking noisy.
Also the nature of the noise is important. Fine grained random luminance (B&W) noise is less objectionable (more "film-like" grain) than coarse or blotchy coloured noise. Pattern noise (harmonically-related electronic noise/junk/hash contamination, particularly likely to occur in a CMOS sensor), causes regular horizontal and/or vertical bands in the noise, and is objectionable in very small amounts, greatly reducing the potential to boost shadows. This Sony CMOS sensor is very good in having clean, low-level noise.
There is an argument that 13-14 stops DR is unattainable in the real world. Lens flare and internal camera & lens reflections are claimed to limit real world DR performance to 8-10 stops. See:
http://forums.dpreview.com/forums/read.asp?forum=1036&message=21939993
Dan.