Problem with tivtc (two pass script, first time i see this)

[OvejaNegra]

normally i use a script like this for the first pass:

#source
MPEG2Source ("dgp.d2v",info=3)
#for the dot crwal (it helps with the decimation and matching)
checkmate(tthr2=4,thr=2)
tcomb(mode=1,fthreshL=6,othreshL=6,fthreshC=30,othreshC=30)
#some times i get better results using both filters this way


#credits cutting
trim (0,46913)

#matching
#mcroma false because the video has too much chroma noise

tfm (micout=0,display=false,mode=0,pp=1,slow=2,mChroma=false,chroma=false,hint=true,cthresh=1,blockx=16,blocky=16,MI=80,metric=0,d2v="dgp.d2v",trimIn="trim.txt",output="TFM_matches.txt")

Crop(12, 0, -8, -0) #noise on borders

#decimate
tdecimate(display=false,chroma=false,nt=1,mode=4,output="TDecimate_metrics.txt",denoise=true,blockx=32,blocky=32,hint=true)


i run this script with megui (analysis pass)

now the second pass script:


#source
MPEG2Source ("dgp.d2v",info=3)

#color correction
#colormatrix (interlaced=true,hints=true,threads=0)
#
#dot crawling (this time is for removing the noise forever)
checkmate(tthr2=4,thr=2)
tcomb(mode=1,fthreshL=6,othreshL=6,fthreshC=30,othreshC=30)

#
#credits cutting
trim (0,46913)
#
#
#clip 2 for external deinterlecer
interp = last.nnedi(field=-2)
dt=last.tdeint(type=2,edeint=interp,mode=2,full=true)

#
#
TFM (micout=0,display=false,mode=0,pp=2,slow=2,clip2=dt,hint=true,d2v="dgp.d2v",trimIn="trim.txt",ovr="TFM_ovr.txt",input="TFM_matches.txt",batch=true)



tdecimate (display=false,mode=5,hybrid=2,vfrDec=0,chroma=false,input="TDecimate_metrics.txt",tfmIn="TFM_matches.txt",mkvOut="tiempos.txt",ovr="TDecimate_overrides.txt",batch=true)

Well, most of the time i use something like thie for my animes.
Now, using this precise script i made the first chapter of Bible Black (everything OK)

But the second chapter, well after running the first pass, the second one has interlaced frames randomly, bat matches and is a total mess.

By deleting filters, i was a ble to determine the culpirit :TFM
Is behaving very wierd, if i enable the display stats on the first pass and on the second pass to compare results, for example the frame 486 matches with C on the first pass detection.
BUT on the second pass it shows 486 matches P (OVR).

simply deleting input="TFM_matches.txt" on the second pass TFM line fixes the problem.

What's wrong?
The frames are changed from one pass to another?

Thanks