We implement a procedure-based on the Wishart-Laguerre distribution-recently outlined by both K. Życzkowski and the group of A. Khvedelidze, I. Rogojin and V. Abgaryan, for the generation of random (complex or real) N × N density matrices of rank k ≤ N with respect to Hilbert-Schmidt (HS) measure. In the complex case, one commences with a Ginibre matrix (of normal variates) A of dimensions k × k + 2(N − k), while for a real scenario, one employs a Ginibre matrix B of dimensionsThen, the k × k product AA † or BB T is diagonalized-padded with zeros to size N × N -and rotated by a random unitary or orthogonal matrix to obtain a random density matrix with respect to HS measure. Implementing the indicated procedure for rank-4 rebit-retrit states, for 800 million Ginibre-matrix realizations, 6,192,047 were found separable, for a sample probability of .00774006-suggestive of an exact value of 387 5000 = 3 2 •43 2 3 •5 4 = .0774. A prior conjecture for the HS separability probability of rebit-retrit systems of full rank is 860 6561 = 2 2 •5•43 3 8 ≈ 0.1310775 (while the two-rebit counterpart has been proven to be 29 64 = 29 2 6 , and the two-qubit one, very strongly indicated to be 8 33 = 2 3 3•11 ). Subject to these two conjectures, the ratio of the rank-4 to rank-6 probabilities would be 59049 1000000 = 3 10 2 6 •5 6 ≈ 0.059049, with the common factor 43 cancelling. As to the intermediate rank-5 probability, application of a 2006 theorem of Szarek, Bengtsson and Życskowski informs us that it must be one-half the rank-6 probability-itself conjectured to be 27 1000 = 3 3 2 3 •5 3 , while for rank 3 or less, the associated probabilities must be 0 by a 2009 result of Ruskai and Werner. We are led to re-examine a 2005 qubit-qutrit analysis of ours, in these regards, and now find evidence for a 70 2673 = 2•5•7 3 5 •11 ≈ 0.0261878 rank-4 to rank-6 probability ratio.