Hg spectral lamp:

The line at pixel no. 1340 is mercury's emission at 546.0 nm. At pixel no. 598 is the 435.8 nm line.

It's not entirely accurate to assume a linear relationship between wavelength and pixel no but let's assume that tan θ is linear for small intervals (and here we're talking about a difference of 6.4° of θ for the diffracted light for the two lines). That gives 0.15 nm / pixel.

Of course I'm not getting a resolution like that.

Here's a capture of a Na spectral lamp:

It's supposed to be two lines at 589.0 nm and 589.6 nm. If you really want, you can sort of persuade yourself that there's a hint of two lines here. But they are really not resolved.

Maybe it's the focusing, maybe it's the plastic grating, maybe it's my sticky fingers on the CCD. Whatever it is, it's smearing the picture.

Let's go back to the CFL-test of the weekend:

With the "calibration" with the Hg-lamp we can now say that he two lines at around pixel no. 1325 and pixel no. 1352 are 4.0 nm

apart. (I'm not sure I can afford the .0 in this statement). But it's safe to say that the resolution of the spectrometer is somewhere between 1-4 nm.