Actually, this is not true any more. Posits (aka Unum III) are (tapered) floating point representations that use 2's complements for floating points. While most implementations (and there are not yet so many) convert this to a signed exponent and unsigned mantissa, some also make use of the two's complement format for faster calculations.

Also the even more recent Takum format (using a base "e" instead of binary for the exponents) uses 2's complement encoding.