At high Reynolds numbers, the logarithmic range in wall-bounded flows spans many scales. An important conceptual modeling framework of the logarithmic range is Townsend's attached eddy hypothesis (Townsend, 1976, The structure of turbulent shear flow), where high Reynolds number wall-bounded flows are modeled as assemblies of space-filling, self-similar, and wall-attached eddies. Recently, Yang et al. (Yang, Marusic, Meneveau, 2016, Phys. Rev. Fluids, 1, 024402) re-interpreted this hypothesis, and developed the "hierarchical random additive process" model (HRAP), which provides further insights into the scaling implications of the attached eddies. For example, in a recent study (Yang, Baidya, Johnson et al. Phys. Rev. Fluids, 2(6), 064602), the HRAP model was used for making scaling predictions of the second-order structure function (u i (x) − u i (x))(u j (x) − u j (x)) in the logarithmic range, where ui's are the velocity fluctuations in the ith Cartesian direction. Here, we provide empirical support for this HRAP model using high-fidelity experimental data of all three components of velocity in a high Reynolds number boundary layer flow. We show that the spanwise velocity fluctuation can be modeled as a random additive process, and that the wall-normal velocity fluctuation is dominated by the closest neighboring wall-attached eddy. By accounting for all the three velocities in all the three Cartesian directions, the HRAP model is formally a well rounded model for the momentum-carrying scales in wall-bounded flows at high Reynolds numbers.