The reason its not seen as clear cut is when research has been done such a change changes a whole host of other behaviours of people e.g. where they choose to live or even how they travel.
For example, a policy that allows hybrid working or fully remote working might lead a portion of employees to move from a city centre where car ownership is low to a suburb where it is high. So you might replace a 5-day a week short commute by public transport with a 2-day a week long commute by car which would generate more emissions. This is more than just a hypothetical and has been observed in some cases.
It’s also worth just noting that whilst digital infrastructure at current levels is usually less carbon intensive than any amount of carbon intensive travel it does have a cost and that the trajectory to more and more intensive technologies is increasing that impact (e.g. blockchain and modern AI techniques)
Lastly, there are efficiencies of scale for heating and cooling that might be achieved in offices which might outweigh the transport costs. This is true where I am partly because offices have been brought up to modern spec by regulation where housing has been let go: being more draughty and less insulated.
Personally, though my take is that whilst these second order effects are super important to look at (since in the short term will be linked to real world emissions) I think they are probably best thought of as ways of showcasing issues in other sectors that need tackling serpately (e.g. the suburbs needing to transition away from carbon intensive travel and land use policies to ensure that we don’t lose the necessary density of our urban environments).
The only time I think it would be important as an assessment of a particular policy is when some cost is intrinsic to that change. Say, for example that the only way home working could function for a particular use case was by using some sort of energy intensive block chain system for authenticity and the additional emissions costs outweighed the benefits of avoided travel.