Unable to reduce fitting error less than 7.2%

[Chemical-Freak]

Dear Dr. Tadano,
I am calculating phonon dispersion of 2D ISbTe monolayer with 5x5x1 supercell. The phonon dispersion plot contained no imaginary modes. However, the harmonic force constant fitting error is 7.2%. I am clueless on how to reduce it further. I used mag = 0.01A. When I used mag=0.02, 0.03 the fitting error only increased.
I am using quantum espresso for DFT calculation, optimized the structure with etot_conv_thr = 1d-06, forc_conv_thr = 1d-05, I have attached alm input and output files. How can I reduce the fitting error? Or is it acceptable?
Thank you.
files.zip

[zohansyahfatomi]

How many k-points do you use for the force constant calculation?

[Chemical-Freak]

Hi zohansyahfatomi! I am sorry for this late reply. I used 3x3x1 k grid size for force sets calculation since I was using 5x5x1 size supercell. But I proceeded with 7.2% error anyhow, since my results matched! But I couldn't figure a way to fix it.
Thank you.