Conjugations are just that... Conjunctions (k I keep getting that word wrong).
And it means just that... there are duplicate entries in it.
[ In all truth, not too sure the duplication is really a problem, just means there are fewer 'so fast that', etc. ]
/* ---------------------------------------------------------------------------------------------------------------------------- */
/* Chart 08 -- Conjunction Phrases */
/* ---------------------------------------------------------------------------------------------------------------------------- */
var xmtab08 = new Array(
"and so it", "before it", "for it", "and in addition it", "so fast that it",
"so it", "then it", "when it", "where it", "while it",
"and", "and", "and", "and", "and",
"and", "and", "and", "and", "and",
"and", "and", "and", "and", "and",
"as it", "as it", "as it", "as it", "as it",
"as it", "as it", "as it", "as it", "as it",
"as it", "as it", "as it", "as it", "as it");
Clustering: psuedo-random behavior that becomes obvious when multiple random calls are made within the extremely short timeframes (very much obvious with the hand gen functions). This has two results, typically that a given hand will have no specials at all, or result in 2 or 3 specials right around each other. How I addressed this with MeowthBot and the test-gen is simple... I changed the fixed random numbers that the specials check against so they shouldn't suffer as strongly from a 'cluster'. Fixing this for table/list field entries (styles, locations, modifiers, etc) is not as easy, and would require a minor shift in general layout of the generator code (notably moving the style, etc tables to be globally declared and shift the values with each move-gen).
One oddity at least... which I also mentioned in my post, blood of the sun, 'ell blood of just about of the structures. Simply put a few of the nouns (defensive targets as I said earlier) are too specific for all the descriptors. Easy enough to fix by either removing the ones that cause the blatant 'err what' factors, either by adjusting the nouns table or the descriptor (okie I screwed up here

, that is actually called the defensive target table, but not really all too appropiately imo but oh well).
For your question, yes, and I actually did mention I had an idea of how to approach the clustering problem with the listfields and will probably start work on that over the coming week(aspects of a move that come from one of the lists or tables, not strictly 'is there a special or not' type things... that is easy to account for).
Another minor change that was made to test-gen at least...
I formatted it so it's more readable to see and understand at a glance (at least in my opinion). This has no actual effect on the generator code-side except being easier to read for terms of making changes if necessary and for looking at the code in these discussions.
[ and now it is my sleep time ]