It isn't accurate at all because with very low fees you can definitely have spam but you can also have transactions that are spewed as manipulation to make the coin seem used.
That would be if this metric was already widely used, and so coins were trying to manipulate the metric. So for the moment, using transactions instead of fees is a reasonable proxy as a first approximation until someone can compile the data for fees.
(Even high fees would be irrelevant if the insiders also control the mining, but this wouldn't generally correlate with a high hash rate.)
That is why I include the hash rate in the metric and then take the square root.
Bytecoin is a good example of this. It has tiny fees and junk transactions generated by some automated process 24x7.
I hope someone can compile the fees data. But we can already see that Bitcoin is 100X ahead of all the altcoins. We can presume the altcoins are generating more spam transactions than Bitcoin, because they have an incentive to hype as you indicate for Bytecoin.
But then, a lot of exchange volume (especially in China) is also documented to be fake, so it very hard to make these comparisons meaningful.
Fees aren't fake in any case (unless very low hash rate and can pay to themself as miners), which is the point of using fees (and multiplying by the hash rate).