Core Image Colorspace?

I've made a macro utilizing the CIToneCurve node to offer similar functionality to what you would get out of levels in Photoshop. More recently, I decided to try to write a CIFilter node to do the same thing. I THINK it worked, but the results differ between the two.

The one I wrote in the CIFilter gets darker much faster, but aside from that, the results are pretty similar. The details for the CIToneCurve node say that "The curve is applied in a perceptual (gamma 2) version of the working space." The Core Image Filter makes no mention of color space though, so I'm attributing this difference in results to the Core Image Filter calculating in linear light space, but I'm not sure. However, if this is true, I think it may be worth going back and replacing the tone curve macro with the CIFilter in some of my previous projects.

Would anyone be willing to step in here and confirm my suspicion?

I've attached a reference Quartz Document with an exploded version of the aforementioned macro, the CIFilter, and a multiplexer to quickly compare results between the two. (feel free to use any resource contained within any of your own projects)