Any LUT has a determined input and output color space, so the footage and display should always match those. If the footage doesn't match you should transform before the LUT, if the display doesn't match you should transform after the LUT, and of course if none of them matches it should be transformed before and after the LUT. In a color managed system with an intermediate working color space such as ACES or RCM2 the input and output transforms are already there, so you have to compensate for those before and after the LUT. There's this post about empirical LMTs that covers this:
https://community.acescentral.com/t/lmts-part-2-how-do-they-work-and-how-are-they-made/1203