For starters, a process (or more) may have been running, say polling emails, backup info (*WHATEVER*) on the lower score one, or not but there may have been which can interrupt the benchmarking software hence giving lower scores. Ultimately, benchmarks aren't a great way to get a 'definite' source of a problematic device, especially on a single run. Multiple runnings, averaging the scores (bell curve, remove highest and lowest scores which may seem 'out' of the normal score), mean and the like to try to guage if one install is better than the other. Plus since the environments most likely aren't the same, besides the obvious 4G thing, it would be hard to see a somewhat/exact similar score.
Moral, for me anyway, unless a device is obviously slower, say frames skipping etc in gaming, things taking lot longer to load than another 'same model' device (and ROM), benchmarks are just an indication of a possible experience. Take with a grain of salt, same as advice over the internet from people when you have no idea of what their qualifications for giving said advice are.
Do the tablets feel different, placebo possibly included since you know one has a lower/better score than the other.