Analysis : Monte Carlo uncertainty analysis with SimaPro using Lognormal distribution (defined by the pedigree matrix for most  inventory items).

I'm getting a very large negative variations (outliers) in the resultant uncertainty range for 'Toxicity' categories (human & ecotoxicity). It is a simple analysis with 7-10 inventory items. The other 2 damage categories (Ecosystemsn and Resources) getting acceptable variations. However, the large negative variation affects the final results giving large std deviations having negative impacts within the 95% confident interval. 

Any idea why negative impact figures ?  or suggested reading, please.

Thank you in advance for your  efforts in responding.