There may be instances where the event data you want to upload contains multiple ObjectIDs for a single event (e.g. Add to cart). There is no standardised way to nest data like this in a CSV file, so how can you update your Recommend model with this kind of event structure?
The solution is to duplicate the row, with only the objectID changing, for each objectID that you are looking to include. It is worth noting that each row in the CSV will be treated as an individual event when feeding the recommend model. However, for models such as Frequently Bought Together, we're looking for things that are bought together over a period of time (24h), so splitting events into multiple rows won't effect the model.
It is also worth noting that the model doesn't make use of conversion sub-types which you may differentiate in your data. It only uses type; "click" or "conversion", so when uploading a CSV the assumption is that events that are not significant are excluded.