You could simplify your datamodel if you don't split your shipping-data for the different transport_mode. There will be no advantage to split them and then creating complicated link-table structures to get them associated again. Keep it simple with one fact-table and if you then joined all measure-fields to the fact-table and keeped the other fields within associated dimension-tables - it will be probably work quite smoothly.
Sorry, I couldn't build this datamodel for you. Without access to the data and a good understanding of the kind and quality of the data and the requirements which data-insights should be realized with them is it not possible to build a reliable datamodel and to gain these knowledge and to do the work needs more efforts then I could here provide and which would be also out of the scope of these community - to help others but do not their work.
My tip for you is that you build the datamodel step by step - at first the main-table (not splitted) and are there further fact-tables concatenate them and afterwards comes the joining/mapping of measure-fields to the fact-table and the other fields are placed in dimension-tables which are associated with the fact-table (is the dataset rather small the further measure-fields could be left within the dimension-tables - the performance-penalty of this would be lesser then the easier handling and logic within the datamodel).
