I have attached the complete bundle which is the ADa662*.json file. This is the one I uploaded via the Medplum batch upload interface, and encounter resource (among others) is not being extracted.
To debug, I extracted small snippets of this file (ada1.json) and compared with lucien1.json which actually works. I think lucien is the sample you provide as one of your example samples, though I believe it is also from Synthea.
The only difference between the snippets seems to be the ordering of the resourceType field within the resource like so:
"resource": {
"resourceType": "Encounter",
"id": "2fae0e14-5528-1dc9-e682-5bd60bf7e018",
for the ada patient vs. a slightly different ordering for the lucien patient.
Ordering of key value pairs shouldn't matter for JSON parsing, so I am not sure if this is the cause. I didn't compare the entire JSON files for the 2 patients. Lucien works, Ada doesn't.
To clarify, this is a slightly different (though perhaps related) issue to the problem of other resources like Claim etc. not being extracted. I was surprised that even Encounter resource is not being extracted for some of these uploads. Claim, Account, Invoice, ExplanationOfBenefit etc. are not always present in many of the Synthea samples ; for example, your Lucien sample, I don't believe has these billing type resources, so they wouldn't be extracted.
Do you expect the FHIR JSON upload to conform to any particular format, besides the standard R4? I assume that any Synthea FHIR R4 should work.
This Ada* json file is from
https://synthea.mitre.org/downloads , one of the 100 samples for the latest synthea version FHIR R4.