Two things to note here. First, subtracting inflation from the nominal interest rate is an approximation to the real interest rate, but only in discrete time. Furthermore, the "true" relationship it's approximating isn't division of one rate by the other--you have to add 1 to all three of your quantities (inflation, real interest rate, and nominal interest rate) first to get the true relationship.
Here's a brief overview. Consider the Fisher equation of
$r = i - \pi$
where $r$ is the real interest rate, $i$ is the nominal interest rate, and $\pi$ is the inflation rate.
This equation is often introduced as a linear approximation to the true real rate of interest, given by the equation
$\frac{1 + i}{1 + \pi} = 1 + r$
Let's see how this holds in a discrete time model. Denote your nominal income as $Y$ and the price level as $P$. Your real income is $Y/P$. If all this income is invested in some interest-bearing asset in a discrete time model, your real income in the next time period becomes
$\frac{Y (1 + i)}{P (1 + \pi)}$
and if we want to find a real interest rate that summarizes this change in real income, we would need to write your real income in the next time period as
$\frac{Y}{P} (1 + r)$
which gives us the identity $\frac{1 + i}{1 + \pi} = 1 + r$ that the Fisher equation approximates. However, if we work in continuous time, this breaks down. First, the units don't work out--inflation rates and interest rates are measured in percentage change per year (or some other unit of time), so they cannot be added to the dimension-less number $1$. Second, it turns out that Fisher's approximation is actually completely correct in continuous time. Using derivatives, we define our quantities as follows:
$i = \frac{dY}{dt} \frac{1}{Y}$
$\pi = \frac{dP}{dt} \frac{1}{P}$
$r = \frac{d(Y/P)}{dt} \frac{1}{(Y/P)}$.
Using the quotient rule, we can rewrite $r$ as
$r = \frac{\frac{dY}{dt}P - \frac{P}{dt}Y}{P^{2}} \frac{1}{(Y/P)}$
which simplifies to
$r = (\frac{dY}{dt} \frac{1}{P} - \frac{dP}{dt} \frac{Y}{P^{2}}) \frac{P}{Y} = \frac{dY}{dt} \frac{1}{Y} - \frac{dP}{dt} \frac{1}{P} = i - \pi$
which gives us the Fisher equation, no approximations about it!
Note that this assumes continuous compounding from your nominal interest rate. If you instead have a compounding rate of $\tau$, we would define $i$ differently, and our equation becomes
$r = \tau \ln(1 + \frac{i}{\tau}) - \pi$.
However, for the purposes of the data you're working with, I am 90% sure that this modification is completely superfluous. If you want to use something more precise than the Fisher equation, you need to know exactly how your data was computed. What price index was used to calculate the inflation rate? Under what assumptions was the nominal interest rate calculated?
In short, while I can't speak to the reasons for your supervisor's recommendation, they're definitely correct that you should subtract inflation rather than divide by it. There's a reason why we use the Fisher equation.
