My expertise is with Video, not stills photography so maybe take this with a pinch of salt but since there aren't many responses yet it might be of value. There's different types of LUTs but I think odds are you're probably thinking of "creative LUTs" or "Look LUTs" so I'm speaking with reference to those.
With video footage shot log, one typically does a transform process first to bring the footage in to colour space appropriate for the project by using a transform function specific to the camera and brand. This is because the LUT will have been designed expecting a specific starting gamma curve and colour space and typically needs to be transformed in to those first in order to avoid unexpected results.
While RAW is a bit different from Log, there are much the same principles at play. It starts out with no gamma-encoding or display colour space but when you open it in editing software, the program applies a debayer and a default color/tone interpretation. In that sense, the “transform” stage happens in the RAW development settings first and any LUTs would necessarily be happening afterwards.
Another good reason to keep a LUT at the end of whatever processing chain you have for your image is because, the way the values are remapped by the LUT means that depending on what the LUT does, you could end up with destructive results were you get clipping or oversaturation and it's best to make sure that that happens last so you can compensate for it ahead of time before you have lost information from this destructive process.