Spectroradiometers are invaluable for obtaining the greatest level of accuracy. However, they also tend to be rather slow, especially at lower light levels. They are also very expensive.
Colorimeters are the exact opposite. They are very fast, relatively cheap, and have great low-light sensitivity. However, they can have problems with accuracy, especially on certain types of displays.
Using a spectro to correct the color response of a colorimeter is, for me, the preferred approach. Others prefer to use the spectro as the sole measuring device.
BTW, it is a myth that spectros are somehow necessary for measuring primary and secondary colors, whereas colorimeters are better suited for grayscale. This claim no doubt grew out of some marketing strategy designed to give consumers a clear reason to invest in expensive equipment. It is similar to the myth that I used to hear often that colorimeters were OK for CRTs, but you needed a spectro for LCDs or DLPs, or LCoS displays. It was another attempt to market spectros by claiming that they were necessary for emerging technologies.
It is unfortunate that these myths have been perpectuated, because there really are good reasons (just not these) for investing in a spectro, especially for pros.
For example, here's the offsets I measured for the Planar between my spectro and colorimeter.
As you can see, the colorimeter did no worse in general with the pri/sec colors than it did with white. These results are typical.