GPT-3 Davinci API not handling large blocks of text well

So I’ve been working with the API to try and generate personal stories based on interview data. The interviews themselves are about 8K - 10K tokens so I’ve been chunking, summarizing and recombining. The final summaries are acceptable but now I need to take that 1K summary and have the API write a story about the subject. Seems I’ve hit a roadblock there. If I get back anything at all, it’s typically a few sentences and it just ignores the bulk of the data. ChatGPT on the other hand more or less gives me what I want if I pass it the same summary data.

For example:

Write a 5 paragraph story about John Doe’s, childhood, education, career, favorite activities based on the following:
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut hendrerit pharetra nulla vitae fermentum. Orci varius natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Cras ornare aliquet lacus. Quisque sodales semper lacus nec commodo. Etiam non leo semper…