I should probably quantify that a little more but let me assure you that these are real values from real colorimeters and they can be that far off from one another.
Spyder 3s from what I've observed tend to report a measured value as being cooler than it actually is. DTP94s by comparison report a measured value as being warmer than it actually is. Eye Ones are somewhere in the middle. If profile my monitor to say D65 with my i1 Pro and the measure the same white point with the Spyder 3 and the DTP94 I will get values of 7400ēK and 5600ēK respectively. The delta is +/- 900ēK in what these pucks are reporting back. They are not to be relied upon for accurate profiling of white point.
ETA - Now these are the differences I've noticed from one manufacturer to another. If you compare the same device to the same device you won't see quite that delta but I have seen 400ēK differences between two devices of the same make and model.
I'm starting to think you may be right (sort of, I do seem to be noticing differences only I'm getting the DTP94b reading things as cooler than they actually may be and the other probes warmer, the exact opposite of what you wrote; EDIT: actually maybe not, it seems that it is perhaps making profiles end up with too much blue and not enough green, anyway whatever the exact details I think you are likely right about DTP94 perhaps not getting white point exactly on spec), at least about the differences in color temp various probes may read.
I just compared a DTP94b calibrated (but not profiled) sRGB CCFL LCD HDTV to a color checker chart (viewed during the day with windows open, but the sun not directly on it, so hopefully that is vaguely D65) and did the same for the set calibrated by the NEC custom puck and the same for NEC PA calibrated with NEC puck and using factory defaults.
The NEC (i1D2) calibrated HDTV looked way red, nothing more needs to be said.
The DTP94b calibrated Samsung CCFL LCD sRGB HDTV and Samsung sRGB monitor actually didn't seem to match quite as closely as the either the NEC PA with SVII+NEC puck or simply using the factory preset.
to be honest none of the displays really matched the checked when it was held up with each color side by side, even the NEC seemed noticeably off, but it did seem to be the closest overall and if the chart was held closeish but not too close it at least looked somewhat ballpark, the other displays needed the color checker held much farther away for things to begin to look ballpark.
Of course I really need to use a D65 lamp to be sure. (then again it seems that ones with a good spectral response cost rather a lot)
(of course you also mention that different software provides different values too and that was certainly an element in play here, very different software and calibration methods)