Bug in SQRT

(sqrt 5.0)
2.236068d0


But from the Hyperspec


12.1.4.4 Rule of Float Precision Contagion

The result of a numerical function is a 
<http://www.lispworks.com/reference/HyperSpec/Body/26_glo_f.htm#float>float 
of the largest format among all the floating-point arguments to the 
<http://www.lispworks.com/reference/HyperSpec/Body/26_glo_f.htm#function>function.



So it must return a single float, not a double float.