Collapsing all the categories will make it difficult to say things like "Using locally sourced construction materials will reduce our carbon emissions by x%," or "Our new SolarShip(tm) technology will reduce cargo ship emissions by y%, which amounts to z% of global emissions."

We don't need to eliminate categories, we just need to be clever in the way we present data. For example, there's nothing particularly difficult to understand in the statement "Cement is responsible for 8% of the world's CO2 emissions, half of which is from transportation, 1/3 of which is from cargo ships." If we likewise break down each category into its major sub-components, the sub-components can be recombined in different ways to produce a lot of interesting and actionable data. Having several charts side-by-side is much better than fighting over a single chart that doesn't add up to 100%.

There was an interesting pilot performed by SEI (Stockholm Environmental Institute) and UCL, using shipping data from Brazil, that tried to attribute shipping emissions to commodity type [0]

0: https://www.sei.org/publications/shipping-emissions-per-comm...

But ultimately the main purpose of such stat is to make policy decisions to reduce amount of greenhouse emissions, and so it should show effects of the both making cargo ships more emmision efficient and reducing amount of things people buy. So it's better to let it overlap.