Accurate　channel　modelling　is　crucial　for　network　performance　analysis,　particularly　when　considering　unmanned　aerial　vehicles　(UAVs).　Measurement　campaigns　involving　UAVs　have　shown　that　the　path　loss　between　UAVs　and　terrestrial　base　stations　(BSs)　is　highly　dependent　on　the　UAV　height,　rendering　previous　line-of-sight　(LoS)　probability　models　inaccurate.　In　this　paper,　the　coverage　probability,　average　ergodic　rate　and　area　spectral　efficiency　are　theoretically　investigated　based　on　a　more　realistic　channel　model,　with　height　dependent　LoS　probabilities.　Our　results　show　that　the　long　tail　of　the　LoS　probability　function　in　UAV　networks　has　an　important　performance　impact　on　sparse　networks　where　the　BS　density　is　low.　Our　new　findings　also　show　that　the　coverage　of　UAV-enabled　networks　may　continuously　decrease　with　the　UAV　density,　since　the　LoS　probability　slowly　declines　with　the　link　range.