Little research has compared land application of stockpiled (SM) or composted (CM) beef feedlot manure with straw (ST) or wood-chip (WD) bedding on loss of reactive phosphorus (RP) in runoff. We conducted a 6 yr (2013–2018) rainfall simulation-runoff study and utilized surface (0–5 cm) soil collected from a long-term (since 1998) field experiment on a clay loam soil in southern Alberta, Canada. The treatments consisted of SM or CM with ST or WD bedding applied at 13, 39, and 77 Mg·ha−1 (dry weight), as well as an unamended control and mineral fertilizer treatment. Surface soil was collected from all treatments after 15–17 (C15, C16, C17; 2013–2015) continual annual applications, and then 1–3 yr (L1–L3, 2016–2018) into the legacy phase after manure applications were first discontinued in 2015. The soil was packed into runoff trays, and flow-weighted mean concentrations (FWMCs) and mass loads of RP5 (<5 μm filter) in runoff water were determined during rainfall simulations. Our findings generally supported our null hypothesis of similar RP5 losses for manure type (CM = SM) and bedding (ST = WD) for most years. Successively higher application rates increased RP5 loss by 32%–121%. Termination of long-term applications dramatically reduced FWMCs by 58%–77% and mass loss by 56%–65% from the C17 to L3 years. This suggests an accumulation of soil P during continuous phase and depletion during legacy phase; therefore, lower application rates or termination of applications may reduce RP5 loss in runoff.