In older forms of the Periodic Table, boron was not necessarily far away from what we now call scandium. Mendeleev himself used a form where what we now call the fourth and later periods were set up in two rows based on how the stoichiometries of compounds tended to repeat (see here). For instance, copper forms an oxide and a chloride with similar atomic ratios to those of potassium, and Mendeleev's table placed copper in the first column with potassium.
[Source: Källa:Dmitrij Ivanovitj Mendelejev (1834 - 1907). This work is in the public domain in its country of origin and other countries and areas where the copyright term is the author's life plus 70 years or fewer.]
However, the columns are staggered as if to indicate that at least in the later periods, two rows were needed to make a full periodic cycle; rubidium (two rows below potassium) bears a closer resemblance to potassium than copper (one row below), despite all three metals forming (for instance) solid monochlorides.
When we move over to column 3 (elements forming sesquioxides and trichlorides) in Mendeleev's format, where the two elements immediately below aluminum seemed to be unknown, the one immediately below aluminum and two rows below boron was thought to be more like the latter, whereas the element two rows below aluminum would be more like that metal. Mendeleev then labeled these expected elements accordingly.
Today, if we look at melting points, we see a sharp rise from aluminum to eka-boron/scandium (which melts at nearly the same temperature as iron), then back down to low-melting (nearly room-temperature liquid!) for eka-aluminum/gallium.