I think I understand, after rereading the post a few times.
20 cycles + 10 cycles of possibly failed or lower efficiency cycles (at a higher anneal temp) may simply be insufficient for a genomic target at the low efficiency expected for a target this large. Assuming the same cycle was done twice in the nesting, that may be 20 + 20 - 10 (dilution) = 30 cycles. For low efficiency conventional PCR this may not be enough. If the higher anneal temp worked well then the cycle number would be plenty, however if you knew that temp worked than there would be no purpose in doing the touchdown to a lower temperature.
You still haven't mentioned the enzyme used. I'll assume that it's a high quality proofreading enzyme, or mix, and proceed. However, if this isn't the case this may be where you should changing things.
Assuming you gave the DNA/RNA concentrations in the final PCR mix (otherwise the concentrations wouldn't be useful to us) you have used 75ng gDNA/25 ul reaction, and 4.1 ug RNA/ 25ul rxn.
A one copy target in genomic DNA will be far more dilute than a gene target in cDNA at a typical level of expression, and will also be more difficult to PCR because the size is longer. However, relative to the size you use less extension time in the genomic than with the cDNA. The extend time may be OK, and may have been more than necessary for the cDNA; I don't know the enzyme and so I can't comment, but you may want to double check your enzyme manufacturer's protocol for extension times.
The anneal temp sounds low, but this is relative to the primers and enzyme mix, which I don't know. I personally would use a single anneal temp close to, but below the primer Tm, unless the enzyme's protocol directs otherwise. The nesting will give you the specificity so touchdown is not necessary. Usually with PCR from cDNA it is not very difficult to achieve a single band, and the fact that you got two bands in the first round may indicate that there is nonspecific priming so a higher temp may work better.
So, I think what I might do would be to do two rounds of nested PCR of 30 cycles each at a few anneal temperatures (primer Tm, Tm-4, Tm-8C, but not less than already worked), and maybe duplicate this with a cosolvent added in case there is difficult template present. For a cosolvent, I currently use Phusion polmerase which comes with a "GC" buffer, and/or some DMSO could be tried (perhaps 3%, this would shift the Tm down about 2 degrees C).