Many sensors collect data in 11 or 12-bits but the data are stored as 16-bit integers. So it is no uncommon to have 16-bit unsigned integer data ranging from 0-2048, or 4096 or similar values. If you convert the data to 8-bit (byte data), then you will lose a lot of the dynamic range of the original pixel values. It is ok to do this but it depends on what your goal is. If you plan on using the data for target detection, change analysis, or any type of spectral analysis, you might want to keep the original data range. If it is a spatial analysis or for publication type images (jpeg), then perhaps it is useful to do. I would use the 'Stretch Data' tool in ENVI to convert the data to byte, which will linearly stretch the original values into a range of 0-255.