I disagree daba.
If the original data is a IEEE 754 floating point number, passed as a bit for bit copy into two consecutive integers "from somewhere else in controller land", and you use a MOV, you will destroy the data. You need the byte for byte, bit for bit, "image" of that float back, and the destination of the COP provides this handy dandy feature.
One of us misunderstood the OP's question...
Yes, you are right, you can use this technique to "reconstruct" an IEEE float passed into a machine via INTs or DINTs, but I didn't see any reference to another machine being involved.
I just attempted to point out that the internal representation in memory bits is interpreted locally.
Receiving data from another machine requires that you fully understand the formatting of the data structure, in the same way that you must understand how each machine stores data of different data types.