Copilot 365 agent does not scan PDF embedded images?

We have a subscription to Copilot 365. I use the app, not the web interface on Windows 11 (upgraded from Win 10). 

I made a custom agent to scan through one of our PDFs. The agent also has 2 other .DOCX files as knowledge. I asked it about a person's contact info. In the PDF the person's name and contact info is in a box. The agent returned the wrong information which is not even in the PDF itself. I have the agent set to ignore outside sources and use only sources I put in. This is a very simple agent.

When opening the PDF in Adobe Acrobat I asked their AI the same question and it found the info.

I suspect that this information in the box is an embedded image in the PDF and Copilot 365 cannot scan the image inside a PDF. Can anyone confirm this? 

We may have more PDFs where we need to get info from an embedded image. Is there a way for Copilot to get this information in the embedded image?

I did a workaround as I have another .docx file as a knowledge source for the agent and I put the person's info as text in the Word doc, and it works that way.