The particular lossy compression used here is nothing to worry about. Here's why.

Camera sensors are more or less linear devices, if you double the amount of light you double the signal produced by the sensor. Our eyes work logarithmically, so if you double the amount of light, it appears much less than twice as bright. Another example of this is if you display pure black and pure white on a computer screen, then find a shade of grey that looks right in the middle of those two colours it will only be around 18% of the brightness of pure white.

Our eyes are thus much more sensitive to slight changes in absolute brightness near the bottom of the scale and almost completely insensitive to small changes in absolute brightness near the white end. The sRGB/Adobe RGB colour spaces (amongst others) take advantage of this and store brightness on a nonlinear scale, using more bits (smaller steps) in the shadows where you're most likely to detect similar shades and fewer bits (larger steps) in the highlights where fine steps go unnoticed.

Nikon's lossy RAW compression maps a linear 14 bit space (that contains values directly from the linear sensor) to a non-linear 12 bit space similar to sRGB etc. in that is uses the bits where they make the most difference.

Technically you are losing data, but it's data that is not perceivable in most cases anyway.