Its hard to say what the reason for the effects you describe could be. What exactly do you mean when you write about "floating point errors"? Is this the wording of a Prime (I suppose you are using Prime and not Mathcad!?) or do you experience what looks like a wrong result?
EDIT: Just noticed your follow up posts which clarify that its a Prime error message.
When you say that cutting the expression and repasting it again cures the problem, one reason may be that you place it at a slightly different position and so its above or below another region which the is reponsible for the different behaviour.
According the plot problem which can be cured by using the Re() function: It often can happen, especially when using more complex and long winding calculations, that, because of round off errors and hard to avoid numerical inaccuracies a result ends up with a very tiny imaginary part (maybe in the range of 10^-10 to 10^-15). With the usual display accuracy which normally is not set to show 15 decimals, you won't see this and the result appears to be a pure real number. Nonetheless the plot considers this a non-real value and will not plot. If all values are of that kind the plot throws the error you describe.
