Chromatography needs sample to be dissolved into a solution (LC) or gas (GC) then compared vs reference std - so that's no good for multiple reasons.

TGA tells you changes in weight when heated - so that's no good for chemistry elucidation. Good for physical performance testing.

Reflective FTIR (or Raman) - can give you good fingerprint info to compare with known compounds. Only real expert spectropists can tell you what the bond vibrations mean as far as molecular composition. So that's no good if you don't have a known reference plastic.

Can you point me to a practical technique that provides the most accurate chemical composition

thank you sir .

There's no one technique, it'll be a combination of techniques and iterative approach getting nearer and nearer to the answer, but only if you have data on known plastics, that was my point. For instance, this lab https://www.aimplas.net/test-types/chemical-analyses/ says they use a mix of techniques. The first photo on webpage shows Raman/FTIR reflective measurement. The spectrum will give an intial clue but only as a comparison to existing known reference spectra for similar plastics. You'll need the spectral library of knowns in the first place.

If you have access, NMR may be a good option as well. There is an excellent book called "NMR Spectra of Polymers and Polymer Additives" by Brandolini & Hills that has (as the title suggests) NMR spectra of loads of different polymers. They also talk about sample preparation and NMR conditions.

https://unitechlink.com/data-analyzer/