In my opinion the ideal course of action would be to characterize each transcriptome independently and establish groups of genes that appear to express at different levels within the genomes.

Then compare these groups of genes between one another either by similarity of their sequences or some their known functionality.

The other options are far less attractive because aligning data from one species to the the genome of the other may produce more severe artifacts.

That being said most of the time success in publishing the results depend on the strength of the results and your ability to come up with an interesting finding, so if one method does not seem to work explore the other.

I think for your first question you follow Istvan Albert

For cross taxa mapping, you can

Map the reads on both genomes (two taxa of nematode)

Take only those reads for further analysis which mapped on both genomes, for the estimation of gene expression in their respective species.

Maybe you consider uniquely mapped reads or one allignment per read which are mapped on both genomes

This way the biasness of insertions and deletions of sequences between the genomes are removed and moreover, you get orthologous regions from your cross-taxa mapped reads.