While random system generation works well for individual worlds, it is a poor tool for 
creating a coherent, consistent, realistic background on the scale of the Official 
Traveller Universe. Population distributions, trade routes, etc. are *not* random in 
the real world: they follow predictable relationships. A referee should "cook the books" 
to ensure that something like these relationships hold, though the results should still 
be possible within the random system.

Take Atlas of the Imperium as a starting point. The data it contains are fairly simplistic:

World type: water present, desert, asteroid belt.
Gas giant: yes/no
Naval base: Imperial, Imperial depot, non-Imperial
Starport type: A, B, C, D, E, X
Scout/secondary base: scout, way station, non-Imperial military or non-naval
High Population: yes/no
Allegiance: Imperial (various), client state, non-aligned, various non-Imperial

It is not that hard to use these data to create realistic UWP's. For example, you could:

(1) Start with the high population worlds. 
Connect each one to its nearest neighbors (these may be in a different sector) with trade 
routes, using rules like these:

	(a) Shorter legs (J-1, J-2) are better

	(b) Higher starport types (A, B) are better

	(c) Way stations are better than not

	(d) No routes through starport X systems, or starport E systems without a gas giant 
	or water

(2) Assign each high-population world a "hinterland," consisting of all the systems of the 
same allegiance that are closer to it than to any other high-population world (this can be 
determined mathematically). Systems with type X starports count as a distinct "allegiance"; 
isolated worlds of differing allegiance are grouped together to form a "hinterland," 
whether there is a high-population world present or not.

(3) Within each "hinterland," roll 2D for all the systems with water present; these are 
"potentially habitable." On 9+, the mainworld is T-prime, on a 12 it is T-norm; all others 
(including desert worlds and asteroid belts) are "hostile." Count up all the habitable 
(T-prime and T-norm) worlds, and distribute populations of 5-8 evenly among them, with 
preference going to T-norm worlds, higher starport types and worlds on trade routes. 

(4) Assign "hostile" worlds on a trade route or with a base present a random population of 
1-6 (1D); all other "hostile" worlds get a population of 0-4 (1D-2).

(5) Place Xboat routes (J-4 or J-6) along trade routes, to connect (in order):

	(a) way stations

	(b) capitals (if known)

	(c) depots

	(d) naval bases

	(e) high-population worlds

(6) Reverse-engineer UWPs to fit the data generated. T-prime worlds have Size 5-9, Atm 4-9, 
and Hyd 3-9; T-norm worlds have Size 7-9, Atm 6-7, and Hyd 5-7.

(7) Adjust the results to conform with canon, where it exists.

This is just one possible method to create a more-or-less realistic population distribution; 
there are surely others. Different assumptions will result in different rules, of course -- 
this should be used to make non-human sectors distinctive, for example.