The problem is most like a result of working with sources that are too small in raw pixel dimensions.
If you have a slide for a 1080p resolution screen, then you have an image that is 1920x1080 pixels. The optimal print size for this would be about 6x3 inches: the image dots (px) divided by the typical for-print resolution of 300dpi. If you have a mockup that takes up less than the screen and you want it 11 x 8.5 or something, then now you are stretching after export and this is Bad For Quality.
The way DTP software (inDesign, QuarkXPress, Scribus etc.) works with this is by "linking" the images and then displaying a screen-only low-pixel-count version for us to evaluate placement etc. When the file is exported to PDF or sent to a RIP for plates, the software refers to the data in the linked file rather than the placeholder image.
In Keynote/Powerpoint etc. the image is usually imported and placed, and destructively downsampled to the low-res placeholder. When you export, you are now exporting data from the low-res copy. If you export one-to-one that is fine, but if you now want to print it, well, you then have to re-add pixels and the quality suffers.
I don't use Keynote, but a quick google shows me people speaking about, for instance, disabling downsampling and/or avoiding downsampling bugs by placing a PDF.
Additionally, if you have type, you want to export as PDF and look to ensure that font embedding is enabled. Type is usually rendered above 300dpi. In Ye Olden Days, black type would be rendered at 1200+dpi and the CMYK randoms would be pasted up.
See ( https://discussions.apple.com/thread/5902465 )