Curiosity got the better of me, and I took a look at that 1871 census. It is, as suggested, a complete red herring.
Apart from the occupation of 15-year-old William, the writing is clear enough for me to read the relevant (or, as it turns out, irrelevant) entries as
Thomas Robson Lodger Unm 22 Accountant's Clerk Selkirkshire, Selkirk
William Robson Lodger 15 Apptce Gar....maker Selkirkshire, Selkirk
This matches two of the sons of William Robson and Euphemia Hogg, who were listed in the 1861 census at Mill Street, Selkirk, with (among others) Thomas, aged 12 and William, aged 5.
What's more, the original spells James and Catherine Deas in the orthodox manner, and Daniel's surname is Caroline not Cassalina. Robert Deas has a middle initial K; James Deas jr's middle initial is H, not St, and John Deas' middle initial is Mck not Mcke. Finally, the schedule numbers reveal that Daniel Caroline and the Robson brothers, although described as lodgers, were regarded by the enumerator at least as households separate from that of the Deas family.
A truly classic example of a thoroughly misleading transcription. I suppose it was from Ancestry? Clearly not from FindMyPast which was quoted above, or from SP, and FamilySearch also has Thomas and William Rubson. FreeCEN haven't started on the 1871 for Edinburgh yet.
Always check the original.