I’m a proponent of detaching though I understand your point about design continuity. If running analyses throughout the design process is the admirable plan then I would agree – staying in the main model would be ideal. I detach in most of my studies b/c I’m constantly renaming materials (“CIP_25FA_4KPSI” for example would be cast in place, 30% fly ash, 4K PSI concrete) so that I can keep track of custom concrete mixes, different metal panel types, etc. I also often bind in the structural model and break it all down into worksets and materials that make sense to me. I would suggest, if you are not the project architect, having a meeting with the team regarding naming conventions for the materials and other possible minor changes that may come up from Tally. If you don’t change the appearance for rendering or the like, you should be able to come to an agreement where you can operate within the main model fluidly. And of course, if you’re in the design phase and doing comparative analysis (“if I change the structural bay size, how would that affect my embodied gwp?” for example) I could definitely see you staying in the central model – it’s just that the hard numbers may not be correct b/c you haven’t gotten into the concrete mixes and other items with more specificity.
As for methods to keep material mappings…you could email the support folks at Tally for a thorough breakdown. I too have had some odd experiences. The material mappings should be saved & synced to the central model and generally that’s been true. However I’ve definitely had models where somehow the connection gets lost and the mappings are gone. It’s been awhile since that’s happened – perhaps after I’ve upgraded Tally and opened an older model? I have also at times (against the IT department’s advisement) kept Tally and WBLCA models open for days to make sure that connection stays up. That’s another reason that in my detached models I try to get specific about my naming conventions for as many materials as I can – so if that happens I can remap with a bit more ease. Additionally, whenever I work a WBLCA, I always have a Word document open on the side and record my assumptions so that I can revisit them if need be. Things like – mechanically attached versus adhered roofing or the amount of rebar I’m choosing in concrete (“slab moderate” or the like) – all those assumptions in Tally that I admittedly don’t always know the answer to and may not be able to track down. They may not all be correct but as long as I stay consistent between a baseline model versus a design model (for, say LEED compliance), I think it works comparatively.