>> fprintf('%.0f\n',av)
602214085699999950000000
>> fprintf('%.0f\n',av_1)
602214085700000020000000
because X*B^P performs two numeric operations on three floating point values and accumulates floating point error with each operation. XeP generates the value directly, without any numeric operations.
Because of this:
Or any of the thousand other threads explaining floating point errors and how they accumulate during numeric operations.
You might also be interested in trying James Tursa's excellent FEX submission:
