I wonder if you could re-assemble the broken (due to decayed bonds) strands to the most likely sequence, given DNA sequence statistics similar to n-gram language models for human language (= languages with an alphabet of more than 4 letters).
https://www.genome.gov/genetics-glossary/Shotgun-Sequencing
You cannot unfortunately do this for dinosaur DNA. The longest human chromosome has 250M basepairs.
With a half life of ~500 years it only takes about 14K years for the chromosome to break down into individual nucleotides. 2^(14000/500) is roughly 170M