With many 50kw AMs reporting reliable HD coverage extending barely 25 miles, and penetration of steel-frame buildings already virually nil, reducing the digital carrier injection level is likely to render the HD coverage meaningless. This is especially true at night when the HD has to contend with skywave interference.
Barry McLarnon, feel free to chime in on this, but you calculated that reliable HD decode under typical skywave conditions would require a signal strength of about 36 mv/m, if I recall. Maybe he'll share with us how he arrived at that figure. Assuming that to be the case, even a 6db reduction would seriously compromise HD performance.
Then there are the cases of major AMs with critical tuning networks and very deep nulls which simply will not work with IBOC-AM, without a complete rebuild of the phasor and LTUs. And special cases such as KDKA, where the pattern bandwidth of their stacked "Franklin" NDA tower is so asymmetrical as to make HD implementation impracticable.
If a nighttime "fix" was possible with simple tinkering of injection levels, this would have been accomplished long before this.