I think you are pushing the definition of a DLF a little too far for this application. The DLF is the ratio of the maximum dynamic response to the the response to the same load, applied very slowly (a static load). Add to this that the system is assumed to respond as a single degree of freedom. Systems themselves are seldom have just a single degree of freedom but the way the math works, you can treat the natural modes shapes of vibration as those degrees of freedom - the total response is the sum of all those DOFs or mode shapes. In your model, you are activating three degrees of freedom (3 mode shapes). Each of these modes, alone, cannot account for all the response. The timing and location of the transient load modifies how mush mass is associated with each mode of vibration. If you had ALL the (X, Y, & Z) mass participate in the 4th mode, it wouldn't be the 4th mode, it would be the ONLY mode, a single DOF and your approach would apply.
Again, the 4th mode accounts for only a portion of the total system mass - note the higher participation factor for your 2nd mode.
Your participation factor for the 4th mode is 0.22181. This number includes your DLF of 1.2. If you change your DLF at 57.3 Hz to a different value, the listed participation factor will change to be (0.22181 * (new value/1.2)).
Aside: To simplify your test model, exclude all mass in the Z direction to eliminate your current modes 1, 4 & 6. If you add a guide EXACTLY where your current 4th mode does not move (a "nodal point" in this mode of vibration), you will eliminate your current 2nd mode (the first mode in the X direction). You can then use the control parameter switch to stop the eigensolution after the 1st mode. This will isolate the analysis to your single mode. Even so, all the mass will not participate.