Always use window patterns (with the black border around the color or white pattern) for plasma - do not use full screen patterns. The % does not refer to the window size, it refers to the % stimulus or amplitude of the pattern. For grayscale you will obviously use the 0% to 100% windows. For color (HCFR's primary/secondary combo measure), use the 75% color windows and a 75% white window. HCFR will then calculate it's basic color decoder % errors and deltaE's for each color (more on this later). What source are you using for the patterns - AVSHD, GetGray DVDs or something else?
I did see your post-cal colors (with the wrong white % reading), but I don't think you posted the post-grayscale numbers - both pre and post numbers were the same. Upload the HCFR .chc files - that would be easier.
The blue gamma dip at 39 contrast shows that you are beginning to clip. If you want more light output, use User or Standard mode.
Without using a reference meter, there is no way to be sure that shooting for a lower red value during grayscale measurements will be correct, but if you get your grayscale close to D65 (.3127/.329) and it looks green/blue then it might be worth trying.
I would recommend starting with getting the grayscale right, maybe even resetting the CMS values to 0. Then take a look at the deltaE and % in HCFR's color measures (or use one of the free spreadsheets in the Calibration forums) and adjust the Color control to get the best trade-off. At that point you could use the CMS to try and tweak, and that will require some fine tuning of the grayscale and color again.