Modified resulting sequences in CSV

[EfresBR]

Hello

I am using the package to evaluate potential G4s.

Starting from a fasta file with an entry such as: GGTGGGTAGTTTGACTGGGGCGG

I analyze using
python3 G4Boost.py -f Sequences.fasta --maxloop 20 --minloop 0 --maxG 4 --minG 1 --loops 10 --noreverse --classifier G4Boost_classifier.json --regressor G4Boost_regressor.json

The result is a gff and a csv.
In the gff the results are:
Sequence_1 0 23 Sequence_1_0_23 23 + GGTGGGTAGTTTGACTGGGGCGG

In the csv however, the G4motif is modified and reduced, missing two Gs in the middle of the PQS (the motif goes form a length of 23 to 21).
GGTGGGTAGTTTGACTGGGGCGG --> GGtgGGtagtttgactGGcGG

Is this normal?
Why does this happen? It is changing the sequence.
Is there a way to obtain the G4-pred; G4-prob and mfe-pred of the entire imputed motif? (the 23 length motif and not the 21-long "modified" which is not really what i want to evaluate?)

Also, regarding G4-topology prediction, the algorithm is designed to give the Gs predicted to be part of G-runs (Gs in mayuscules), but not its actual predicted topology (parallel, antiparallel or hybrid), correct?

Thanks for the time

EBR