For one of my projects, I have tested and/or implemented probably six or seven different methods of going from an image to a PDF in the last six months. Ultimately I ended up coming back to ReportLab (which I had initially avoided for reasons similar to those you described) because all of the others had glaring limitations or outright omissions (such as the inability to set document metadata).
ReportLab isn't as difficult to handle as it appears at first glance and it may save you a lot of headache-laden refactoring later on. I would strongly suggest you go ahead and use it and therefore know that if you ever want to be able to do more you will have the ability too rather than do what I did and bounce back and forth between a number of different utilities, libraries, and formats.
EDIT:
It is also worth mentioning that you can bypass the Platypus layout system that comes with ReportLab if all you want to do is put bit a of text and imagery on a page.