Solution
Using un-scaled data versus floating point data does not always necessarily mean the file will be smaller. For example, if the file is a text file and the floating point numbers have less significant digits than the integer data, the floating point file will actually be smaller. This corresponds to the fact that each character in a text file is an 8 bit integer (ASCII), so the more characters, the larger the file will be.
Integer vs floating point is more worth considering for processing performance than for data size.
This DAQmx Help topic explains it in more detail. In general, it is only recommended to use unscaled data if you see performance issues with floating point.
Whether using I16 vs F64 will actually save disk space will be dependent on the exact file type that will be used. TDMS for example is a file type that does not differentiate between these two formats. However, if the file type you will be using differentiates between the different formats, it would probably make sense to use whichever data format that will optimize size with that file type.