@reddeercity:
1) There's no simple route within Premiere Pro CC 2014 to use the ACR grading engine, it's been entirely replaced by Speedgrade and Dynamic Link. Note the "source settings" panel displayed by PrCC2014 when you bring in CinemaDNG footage isn't intended for grading, it's just a way to match the hardware so it looks vaguely "normal" on the source monitor. All the heavy-lifting to color-correct and grade the footage happens afterwards, ideally within Sg (or by creating a Lumetri LUT in Sg and applying it in Pr). If you want to use ACR to adjust your cDNGs, you have to use After Effects (in response to steffenhaldrup you can use Dynamic Link or render to a mezzanine format, it's up to you - Dynamic Link won't give you MPE-enabled playback in Premiere as you're still pulling single frames, but you benefit from real-time updates when you tweak something in AE.
2) You can absolutely use a second display for monitoring - either a second computer monitor or a hardware output device feeding a reference TV. Just open Premiere's Preferences > Playback and tick the box you want under "Video Device". You even have offset fields to account for any loop delay between the audio and video hardware. Be careful when using computer displays fed from a consumer graphics card as not all of them allow individual calibration of each display, so people tend to require physically different cards or a multi-head pro series card. For example the video-out port on a laptop is rarely able to hold a separate calibration from the laptop's own panel.
The change to a Pr<>Sg pipeline takes a bit of mental adjustment; I agree that for our specific workflow it doesn't help much, but ML cDNGs are very much an edge case. The pipelines for CC 2014 are designed with digital cinema projects in mind, where Sg is an established tool - usually in the hands of a dedicated colorist - and the idea of adjusting 'video' footage in Camera Raw is very alien. The new workflow is optimized for true "cinema" cameras (BlackMagic etc), so while the Sg team admitted support for ML-5DIII in their blog it's not an advertized feature, and it's entirely a by-product of support for other hardware. Some flavors work, some don't, but to put it bluntly if you can make your ML files exactly the same internal format to BMCC's footage, it will work perfectly. So far you're close, but not close enough.
Yes, the old ACR > Pr workflow was more flexible, as you weren't limited in what you could feed the frame server - but it caused no end of hassle for pros as it put a bunch of important grading decisions at the front of the pipe. That's not how studios work.