Yes, but since you are losing precision anyway, you can truncate Nat to 64 bits before converting to Float.
For Rust, you may need infinite precision fraction as well, for example fraction - Rust
You can also decode a Candid Nat to u128 in Rust (decoding will fail if the value is more than 128bit). Might be easier to convert u128 to f32.