Claude M365 MCP - Recent GraphAPI update from MSFT

### What happened?

The Microsoft 365 connector's meeting-transcript:///events/{joinUrlToken} resource successfully returns full transcript metadata (transcript IDs, createdDateTime, endDateTime) for a Teams meeting, but the content field on every transcript object is always an empty string (""). The actual transcript text never comes through, regardless of which meeting, date, or time-range scoping is used.

### What did you expect to happen?

Prior to the GraphAPI update, users could seamlessly request transcripts of attended meetings.

The content field should contain the actual transcript text — equivalent to what Microsoft Graph's "get transcript content" endpoint (GET /me/onlineMeetings/{meetingId}/transcripts/{transcriptId}/content) returns, as distinct from the metadata-only "list transcripts" call (GET /me/onlineMeetings/{meetingId}/transcripts).

### Steps to reproduce

Steps to Reproduce:

1. Connect the Microsoft 365 connector, authenticating as a user who is a genuine attendee or organizer of a Teams meeting with transcription enabled.
2. Locate the calendar event (via calendar search), and read it to obtain its meetingTranscriptUrl.
3. Call read_resource with meeting-transcript:///events/{joinUrlToken}.
4. Observe the response: it includes correct meeting metadata and a list of transcript objects (id, createdDateTime, endDateTime) — but each transcript's content field is "".

### Area

MCP Connector (adding/managing servers)

### MCP Server (if applicable)

M365 MCP

### Error messages or logs

```shell

```

### Additional context

This is distinct from a permissions issue. Prior to fixing an identity mismatch (the connector had been authenticated via an admin/service account rather than an account that was an actual attendee of the target meetings), every call to this resource failed with a 403: "Failed to fetch online meeting caused by: 3003: User does not have access to lookup meeting". After reconnecting as the actual attendee, that error disappeared entirely and full metadata now returns cleanly and consistently. The empty-content symptom persists independently of that fix, on an otherwise fully-authorized call.
Related but distinct: #206 (meeting-transcript resource always resolves to the first occurrence of a recurring series) — the same resource area has at least one other known gap. This report is a separate symptom: content never populates, independent of occurrence-resolution correctness.
Suspected root cause (speculative, for triage): Microsoft Graph splits transcript retrieval into two calls — "list transcripts" (metadata only) and "get transcript content" (actual VTT text, at a /content sub-path). It's possible the connector's meeting-transcript resource only completes the first call and never issues the second.
Workaround confirmed: calling Microsoft Graph's get transcript content endpoint directly (e.g. via Graph Explorer, same delegated user/tenant) returns correct transcript text for the same meeting/transcript IDs that return empty via the connector — indicating this is specific to the connector's implementation, not a Microsoft Graph API or tenant-permission constraint.