I am making a guided de novo annotation with Stringtie. I follow a standard Stringtie workflow:

1) Making guided annotation for all samples

for i in "${files_id[@]}"; \
do \
$programs/stringtie-2.2.1.Linux_x86_64/stringtie \
$bam/$i/${i}Aligned.sortedByCoord.out.bam \
-G $human_genome_gtf \
-o $gtf/${i}.gtf \
-vp 15 \
-f 0.1 \
--fr; \
done;

2) Merging annotations and excluding transcripts with FPKM < 1 $programs/stringtie-2.2.1.Linux_x86_64/stringtie \

--merge \
-v \
$gtf/1.gtf \
$gtf/2.gtf \
$gtf/3.gtf \
$gtf/4.gtf \
$gtf/5.gtf \
$gtf/6.gtf \
-G $human_genome_gtf \
-vp 15 \
-f 0.1 \
-T 1 \
-F 1 \
-o $hek/gtf/merged_annotation_filtered.gtf

3) Making annotation for all samples again but with a newly made guided annotation and not permitting for novel transcripts

for i in "${files_id[@]}"; \
do \
$programs/stringtie-2.2.1.Linux_x86_64/stringtie \
--fr \
-vp 15 \
-f 0.1 \
-A $gtf/${i}_merged_annotation_filtered.tab \
-eb $ctab/${i}_filtered.ctab \
-G $gtf/merged_annotation_filtered.gtf \
-o $gtf/${i}_merged_annotation_filtered.gtf \
$bam/$i/${i}Aligned.sortedByCoord.out.bam; \
done;

So I am trying to filter out genes that have FTKM less than 1. But when I plot FTKM , I see a lot of genes with FTKM below 1. How is this possible?