NARS-GPT inferences

[Y A]

Hello,

I have been playing around with NARS-GPT, and I have been trying to see how to use it (with gpt-4-turbo-preview), and I got this:

John said there was a fire
Input: John said there was a fire
<(john * there_was_a_fire) --> say>. :|: {1.0 0.9}
Carl claims there was a fire too
Input: Carl claims there was a fire too
<(carl * there_was_a_fire_too) --> claim>. :|: {1.0 0.9}
John and Carl say they were terrified about the fire
Input: John and Carl say they were terrified about the fire
<john --> [terrified]>. :|: {1.0 0.9}
<carl --> [terrified]>. :|: {1.0 0.9}

Input:
John says he extinguished the fire
Input: John says he extinguished the fire
<(john * the_fire) --> extinguish>. :|: {1.0 0.9}
John says he used gasoline to extinguish the fire
Input: John says he used gasoline to extinguish the fire
<(john * gasoline) --> use>. :|: {1.0 0.9}
<(john * the_fire) --> extinguish>. :|: {1.0 0.9}
Carl says he went to the bathroom
Input: Carl says he went to the bathroom
<(carl * the_bathroom) --> went_to>. :|: {1.0 0.9}
who is lying about what ?
Input: who is lying about what ?
Carl is likely lying about extinguishing the fire, i=9, due to the lowest confidence value of 0.0940136.

I don't understand two things:

1 - Where did the "Carl extinguished the fire" come from ? (red arrow in the screenshot above)

2 - Where does this answer come from:

Input: who is lying about what ?
Carl is likely lying about extinguishing the fire, i=9, due to the lowest confidence value of 0.0940136.

I was expecting I could get : John is lying about the material (gasoline) he used to extinguish the fire 

Am I misusing NARS-GPT? (Just discovered it an hour ago..)