After review of your data, I think there is a strong correlation between the total AI Value of the Tech and the frequency it is awarded.

The following table demonstrates this with the first seven techs.

_____SGs' Occurance________Frequency__________AI Value plus Civ Mod
_____Civ__Neu__Mil____Civ_____Neu_____Mil______Civ__Neu__Mil

Hor ____2__132__169___0.30%__22.45%__33.53%___3___4___5
War____4___23__120___0.60%___3.91%__23.81%___3___4___5
Pot __135___27____0__20.09%___4.59%___0.00%___5___4___3
Cer __171__244__160__25.45%__41.50%__31.75%___5___5___5
Alp __393__237___76__58.48%__40.31%__15.08%___6___5___4
Bro __223__331__342__33.18%__56.29%__67.86%___5___6___7
Mas__171___21____3__25.45%___3.57%___0.60%___5___4___3

Please note that the freqency percentage is calculated against the total number of awarded techs.

For civilized AIs, the most frequently awarded tech is Alp, followed by Bro, Cer, Mas, Pot, War and Hor. This is also the rank order of the Civ modified AI value.
The neutral and militaristic AIs also follow this rule. This also tends to work for the more advanced techs.

Allowing for some kind of historical preference, like Egyptian masonry or Sioux horseback riding, I think it is possible that the modified AI value affects the order in which beginning techs are assigned.

I propose that the initial starting tech is randomly assigned with a preference for techs with a higher modified AI value and that subsequent starting techs are also assigned using the AI values. The menu of subsequent techs is probably limited to what could be normally be researched less the "killer" techs like writing or math.

I would like to look at your complete data set. I think it might show a few more historical preferences as well as show additional distribution patterns.

Congrats on a fine job,

Alphabet: 56
Cer Bur : 27
Bronze : 19
Code o L: 17
Masonry : 17
Pottery : 13
Currency: 3
Mysticis: 1