Complete experimental transition probability density functions P(EЈ,E) have been determined for collisions between highly vibrationally excited azulene and several bath gases over a wide energy range. This was achieved by applying 2-color ''kinetically controlled selective ionization ͑KCSI͒'' ͓U. Hold, T. Lenzer, K. Luther, K. Reihs, and A. C. Symonds, J. Chem. Phys. 112, 4076 ͑2000͔͒. The results are ''self-calibrating,'' i.e., independent of any empirical calibration curve, as usually needed in traditional energy transfer experiments like time-resolved ultraviolet absorption or infrared fluorescence. The complete data set can be described by our recently introduced monoexponential 3-parameter P(EЈ,E) form with a parametric exponent Y in the argument, P(EЈ,E)ϰexp͓Ϫ͕(EϪEЈ)/(C 0 ϩC 1 •E)͖ Y ͔. For small colliders ͑helium, argon, xenon, N 2 , and CO 2) the P(EЈ,E) show increased amplitudes in the wings compared to a monoexponential form (Y Ͻ1). For larger colliders, the wings of P(EЈ,E) have an even smaller amplitude (Y Ͼ1) than that provided by a monoexponential. Approximate simulations show that the wings of P(EЈ,E) at amplitudes Ͻ1ϫ10 Ϫ6 (cm Ϫ1) Ϫ1 have a nearly negligible influence on the population distributions and the net energy transfer. All optimized P(EЈ,E) representations exhibit a linear energy dependence of the collision parameter ␣ 1 (E)ϭC 0 ϩC 1 •E, which also results in an ͑approximately͒ linear dependence of ͗⌬E͘ and ͗⌬E 2 ͘ 1/2. The energy transfer parameters presented in this study have benchmark character in certainty and accuracy, e.g., with only 2%-5% uncertainty for our ͗⌬E͘ data below 25 000 cm Ϫ1. Deviations of previously reported first moment data from ultraviolet absorption and infrared fluorescence measurements can be traced back to either the influence of azulene self-collisions or well-known uncertainties in calibration curves.