Looking at this resource, these data are microarray data rather than RNAseq data. Thus the "signal" in question is the intensity of florescent light being emitted from the the probe associated with your gene. The relationship between intensity and quantity of RNA is different for every probe because each probe will have a different affinity for complementary fragments dependent on the sequence in question (not least the melting temperature).
Affy intensities are usually presented as log values going up to around 12, but the raw number is next to meaningless and usually only means something as a normalized (e.g. RMA) log fold change.
I've never come across RefExA before, and it appears to be undocumented and unmaintained. I wouldn't rely on it for usable data if I were you. A more modern database of expression data is the EBI Expression Atlas. It presents data from a wide range of different studies, processed and normalised on a unified pipeline and expressed as TPMs.