Incorrect results in GPT + Google Sheets integration vs. chat version which produces accurate output

I am using the GPT plugin in Google Sheets for categorization of a list of companies and respective descriptions. I am having troubles with making it return a correct result. I’ve cross-tested with the chat version, and the results are different and much better. My issue is that I can’t use the chat version due to character limit (it’s a long list). Is there a solution for this? Or am I doing something wrong?

Prompt in Google Sheets:
=gpt(“does this company make baby strollers, carriers or baby car seats? Reply yes or no only, please don’t make assumptions. Base your reply on the company description provided”,cell reference)

Example output:
Denmark based company involved in production, marketing and distributions of safety equipments, textiles and furniture products for children . → YES
Baby and infant feeding products supplier. → NO
Dairy manufacturing plant in China. → NO

Obviously the first one is false positive, which is correctly classified by the chat version, but the Sheets GPT lists it as “yes”.

The version used is GPT4. Is there a way to improve accuracy so it matches the chat version? With a list of 500 companies it will take too long to categorize manually or screen for errors.