Rice and arabidopsis diverged more than 130 mya so ks will not reflect time for many of the "pairs" (orthologs) as you mentioned. Ks alone cannot be used to infer orthology or divergent time of such pairs.
even if your question is not totally clear, I see no reason why you could not compute Ks for orthologous genes in rice and arabidopsis (make sure these ortohologs have a one-to-one relationship for more reliability).
To my opinion the main concern is not whether it is correct or not but more what you want to conclude about the computed Ks value. This Ks value might more informative as a relative value (if you want to compare it to other Ks values between other species with similar or different divergence time) than as an absolute value (which will be little informative unless you have some molecular clock data).
If you want some more concrete example you can look at the opossum genome paper where they computed Ks for human-opossum 1-to-1 orthologs and compare it with that computed between human and chicken (respectively 180 Mya and 310 Mya of divergence, the extent of what you can visibly expect between rice and arabidopsis).