JPGs are defined in-camera via a myriad of settings. Would not trust a camera's JPG as an example of what it can do since there's no real way to know how it was configured. In my experience the only way to know what a camera is capable of is to use it and use it for what one shoots in real life. I take any new camera of my forest photo walks and shoot RAW. JPG is a convenience, but even for "soccer dad" shots I still prefer RAW. I have yet to find an in-camera JPG that pleases me. I must be a tough crowd.
I don't think anyone (here) is using the jpg settings to determine what the camera can do... however, I think some folks might be concerned that a regression in jpeg rendering
might be a reflection on a company's overall philosophy... and that could be disconcerting.
* IIRC, Olympus m43 cameras deliberately underexpose the image and then pull the up during jpg processing, in order to save highlights and provide closer to what some people feel is a "filmic" look.
Well, I'm a RAW shooter myself but I know lots of people who shoot mostly JPEG, so JPEG quality IS important.
Anyway, the same comparison page shows that the E-M1 is noticeably WORSE than the E-M5 in RAW, especially at ISO 3200 and up; judging by the low frequency and obnoxious nature of the noise I'd say Olympus decided to follow in the steps of Pentax and employ RAW noise reduction, which is a very bad thing in my book.
I remember there was a bit of a stink about this when the K5 was released. However, DXO testing demonstrated that the high-ISO RAW NR seems to have not adversely affected the output, as the K-5 measured out every bit as good as the Nikon D7000, which uses the same sensor.
The graphic you've posted comparing the M1 vs M5 sensor is disturbing, at first blush there appears to be serious regression in high-ISO IQ. But perhaps it is the EM5's RAW files which have sharpening applied?
* I'd postulate that all RAW files are cooked to some degree, anyway, wrt NR and tonal curve. The nanosecond after light hits the sensor the image data is being interpreted by human decisions. RAW viewing/editing software is interpreting RAW data also, and ACR's interpretation is different than, say, Silkypix's. Basically, as long as the camera behaves as I expect it to, I'm OK with it.