both the quotes I just gave are directly from papers referenced by poore and nemecek.

here’s another

LCA results can have high uncertainties because of the large amounts of measured and simulated data and the simplified modeling of complex en- vironmental cause-effect chains. Recent studies have highlighted the contribution that system as- sumptions and value choices can also make to overall uncertainty (36, 37). A number of quantita- tive uncertainty assessments are available (38) butare rarely used in practice. One of the key questions is, how much uncertainty is acceptable, depending on the application? In some cases, rough estimates of input values can be enough to identify supply- chain hotspots (39), but for other applications, such as product comparisons (37), the demands for more accurate values are higher. For some im- pact categories such as toxicity, very large differ- ences in inventory results are needed to statistically differentiate product systems, whereas for other categories, differences of a factor of two or less may be enough (40). LCA practitioners should al- ways attempt to manage the decision-maker’s expectations and clarify that LCA is not always a tool to provide a single answer, but rather one that permits comprehensive understanding of a problem and its possible solutions.

uh… I’m on my couch on mobile but I have a quote

First, it is often cited that LCA results should not be compared (Desjardins et al., 2012; Foster et al., 2006; McAuliffe et al., 2016; Röös et al., 2013) due to variation in methodology choices, functional units, as well as temporal and regional differences2. Second, no single comprehensive review was identified that adequately covers the breadth of fresh foods available to consumers and caterers. As Helle et al. (2013, p.12643) state ‘data availability and quality remain primary obstacles in diet-level environmental impact assessment’, while Pulkkinen et al. (2015) calls for the creation of a database that communicates data quality, uncertainty and variability to reliably differentiate between the GWP of food types. Previous studies have compiled LCA data to compare different foods (e.g. Audsley et al., 2009; Berners-Lee et al., 2012; Bradbear and Friel, 2011; de Vries and de Boer, 2010; Foster et al., 2006; Nijdam et al., 2012; Sonesson et al., 2010; Roy et al., 2009). While these are useful attempts, the identified studies are inadequate in the coverage of fresh foods available. Environmental Product Declarations (EPDs) attempt to inform consumers of the environmental impacts (carbon, water and ecological footprint) of specific foods, however they also fall short in breadth of items covered at present. The most comprehensive attempt at carbon footprint labelling was performed by Tesco (2012), however failed to label key categories such as fresh fish, pork, lamb or beef before finishing in 2012 due to the scale of the labelling scheme and a lack of participation from other retailers (Head et al., 2013). Third, studies that do compare results may often present singular figures. Peters et al. (2010) and Röös et al. (2011) argue that a range of impacts should be reported from LCA’s to better represent the variety of environmental impacts, as opposed to a singular figure. Finally, there is a lack of synthesised open access LCA data in the public domain available to consumers to inform decision-making.