Hello,
the values kg CO2-eq values given for transport services are from the ecoinvent database, so one would have to look for more detail on how to handle empty truck runs in the ecoinvent documentation.
Transports can be found in the Report No 14 (Spielmann M., Dones R., Bauer C. & Tuchschmid, M. [2007]: Life Cycle Inventories of Transport Services. Final report ecoinvent v2.0 No. 14. Swiss Centre for Life Cycle Inventories. Dübendorf). The system characterisation is described ich chapter 4 (pp. 11-14).
One important point is that you can find two different types of datasets: \"operation, ....\" (unit vehicle kilometres, vkm) and \"transport, ...\" (unit transport kilometres, tkm). The operation datasets describe the running of the trucks, their fuel consumption, and the related airborne emissions. If you choose the transport dataset, you also include burdens from manufacturing, maintenance and disposal of the vehicle, as well as for construction, operation and disposal of road infrastructure.
The vehicle operation is used within the transport datasets with an average load factor, based on statistical data and literature. Thus, the transport datasets (tkm unit) can not be differentiated according to the truck\'s load factor. You might want to consider to work directly with the operation datasets, and apply your specific load factor. However you will then miss out on the burdens from infrastructure (vehicle+road). Another option would be to say that the average load factor used in the datasets is based on all truck transport (empty and loaded), so that actually there is no explicit differentiation between the two.
A good summary on how ecoinvent handles the transport data can be found in
this PDF article: (Michael Spielmann and Roland W. Scholz: Life Cycle Inventories of Transport Services. Background Data for Freight Transport. Int J LCA 2004 (OnlineFirst): 1–10)