I don't think that's a real problem because the pool of dice generated possible private keys is still huge. But I would prefer a dice method that covers the full range of possible private keys. 100 rolls should be enough, but that could produce number that are too large.
Is there any easy solution to this?
not really...
You are getting 255.91 bits of entropy with 99 dice rolls. The idea that you're losing almost 6% of "something" makes it sound much scarier than it is.
I will bet that a private key that has only 224 random bits is still for all intents and purposes secure, even though 99.999999999%+ of the key space is being left on the table.
And then, keep in mind that a bitcoin address only has 160 bits of entropy in the first place.