The paper is a meta-analysis, it's not trying to calculate those things at all. It's collating and standardising the results of other studies doing that. To take Ridoutt et al 2011 as an example because it's the first beef one that comes up in the dataset, regarding feedlots:

In the feedlot subsystem, water and energy use was calculated using data reported in a benchmarking study of Australian beef cattle feedlots (Davis and Wiedemann 2009). The composition of the feedlot ration was based on detailed, multi-year records provided confidentially by a large feedlot operator. Consumptive water use associated with the production of each feed component was calculated using national statistics (ABS 2008a,2008b) and various CSIRO data sources (e.g. Ridoutt and Poulton 2010). The EcoInvent v2 database (http://www.ecoinvent.org) was the source of water use information for mineral supplements (<0.01% by mass). The feedlot operator also provided data on the transportation distances of the feed components which were used to calculate fuel use in transporting commodities to the feedlot.

It's depending on the work of the 1,530 source papers to calculate the inputs appropriately. You would know this if you had looked at the paper, so where did you get the idea that it is as you described?