Low assigned alignments

Basecalls performed using CASAVA version v1.8.2 Trimmed reads with
fastx_quality_trimmer 0.0.13 with a quality treshhold of 18 and a
length of 20 Aligned with Bowtie 2.1.0 and Tophat 2.0.10 using Gencode
v19 junctions Samtools 0.1.19-44428cd to make a bam, sort, index Raw
counts were generated using htseq_count 0.6.1 using the UCSC HG19
known gene transcripts from Illumina iGenomes edgeR 3.2.4 in R 3.1.0
were used to normalize counts between samples and to make comparisons
between groups. Genome_build: hg19 Supplementary_files_format_and_content: Normalized reads in CPM for
each detected gene along with differential gene expression log2 (fold
change) and p-values for each time point

This is the data processing detail for a paper.
I downloaded fastq files from this paper and used STAR alignment and featureCounts to get my read matrix
with the reference of Homo_sapiens.GRCh38 from ENSEMBL. However, with featureCounts, only 3% reads were aligned.
(This is single-end fastq file so I used -s 0, -s 1, -s 2 option for featureCounts but all results were bad)
What went wrong with my pipeline? Is it related to reference genome that I used?