NCBI allows you to pick the nucleotide sequence you would like the gene to be displayed on. See the dropdown list next Genomic Sequence for options.
This decision will determine the context of how the gene is drawn, which affects how the bases are labeled. For example, the numbering typically changes slightly between genome versions, so make sure you know the genome version that your genome was generated against. To see how this works, we'll use an example:
Look at the "Ruler" in the default view of the gene in the default view.
Now, change the Genomic Sequence to RefSeqGene. Look at the new ruler.
In both cases, the top ruler represents the entire sequence the gene is aligned with, and the lower ruler is the region that is scaled so that the gene prediction fills the screen. The same transcripts and proteins are shown, but the scale is different because the sequence they are mapped to. (If you need convincing, look for NM_000041.3 and NP_000032.1 on both screen shots.)
The RefSeq sequence is much smaller than the chromosome assembly, as expected, and the base positions on the ruler refer to those of the smaller sequence. Base what view you use on what data you are trying to find. The RefSeq view might be easier to digest, but most data sets are going to be based on genome builds, so choose wisely.