This is a short-coming of the 0-100% mapping of the color temperature. Unfortunately the absolute color temperature is not linked to the 0-100% in a fixed way. The mapping depends on the capability of each device and you need to find the corresponding values by trial-and-error.
As @J-N-K says, I guess the two bindings map the color temperature slightly differently. IIRC Zigbee maps it between the max and min values that the device reports and probably Hue does something different.
Ideally there would be a standardised type for this - I think this was discussed some time in the past but not taken up. An alternative could be a standard channel type with a standardised definition of what the mapping is, but again I don’t think that exists at the moment (unless I missed it ).
The deconz REST API reports the capabilities of the device in mired, I guess this is directly derived from what the device itself reports. The deconz binding uses this for an absolute color temperature channel.
Even if I doubt that these are exact matches (I doubt that manufacturers calibrate their lights), this is most likely less than with the 0-100% approach.
Again though, while there are lots of different things that CAN be done, unless there is a standard way to do this, each binding will inevitably do something different, and I guess that was the basis of the original question.
From my view it is even more complicated. If i remember it right than the hue binding handles hue bulbs like 0% - 100% equals 2007k - 6000k and if you use a ikea tradfri bulb it is the other may round like 0% - 100% equals 6000k - 2700k