Previous mangrove tree ring studies attempted, unsuccessfully, to relate the δ18O of trunk cellulose (δ18OCELL) to the δ18O of source water (δ18OSW). Here, we tested whether biochemical fractionation associated with one of the oxygen in the cellulose glucose moiety or variation in leaf water oxygen isotope fractionation (ΔLW) can interfere with the δ18OSW signal as it is recorded in the δ18OCELL of mangrove (saltwater) and hammock (freshwater) plants. We selected two transects experiencing a salinity gradient, located in the Florida Keys, USA. The δ18OCELL throughout both transects did not show the pattern expected based on that of the δ18OSW. We found that in one of the transects, biochemical fractionation interfered with the δ18OSW signal, while in the other transect ΔLW differed between mangrove and hammock plants. Observed differences in ΔLW between mangroves and hammocks were caused by a longer effective leaf mixing length (L) of the water pathway in mangrove leaves compared to those of hammock leaves. Changes in L could have caused the δ18OCELL to record not only variations in the δ18OSW but also in ΔLW making it impossible to isolate the δ18OSW signal.