The decibel scale is a relative one, i.e. you can't really express a value \$x\$ in decibels, rather you always take something like \$20 \log(\frac{x}{x_0})\$ for some fixed \$x_0\$. For the usual scales we use, there is a well defined reference \$x_0\$ such as for \$dBm\$ is logarithm of power divided by one milliwatt, or dBSPL is the logarithm of sound pressure divided by 20 microPascals.
LTSpice doesn't know which reference to use, so it apparently always refers to one unit of the corresponding SI unit. So in LTSpice dB for voltage is \$20 \log (V/1 \mathrm{volt})\$ (which by the way gives exactly the value you report for 5 volts). Correspondingly, the current is referenced to \$1\$ ampere, so the dB values are \$20 \log(I/1 \mathrm{A})\$. These tend to be negative, since you don't usually have one ampere currents running through LC circuits when fed with 5 volts.
So yes, it has to do with \$I_{ref}\$, and \$I_{ref} = 1\mathrm{A}\$ in LTSpice. It cannot be changed as far as I know.
This is actually quite clumsy, since one is often interested in transfer functions, which should be referenced to the amplitude or power of the input signal. The only workaround I've found is to always plot a formula where I explicitly divide by the input amplitude, instead of plotting directly voltages in the simulation