bfloat behavior and erf function



Maxima 5.28.0-2 http://maxima.sourceforge.net
using Lisp GNU Common Lisp (GCL) GCL 2.6.8 (a.k.a. GCL)

/* This first erf only needs one bfloat followed by one expand: */

(%i1) bfloat(erf((sqrt(2)*%i+sqrt(2))/2));
(%o1) erf(5.0b-1*(1.414213562373095b0*%i+1.414213562373095b0))

(%i2) expand(%);
(%o2) 4.741476366409942b-1*%i+9.692642119442159b-1

/* but this erf needs bfloat, expand, bfloat to do the job: */

(%i3) bfloat(erf((sqrt(2)*%i-sqrt(2))/2));
(%o3) erf(5.0b-1*(1.414213562373095b0*%i-1.414213562373095b0))

(%i4) expand(%);
(%o4) 1.020646306179612b0*%i*sin(2.5b-1*%pi)
 -3.50100287874768b-1*sin(2.5b-1*%pi)-3.50100287874768b-1*%i*cos(2.5b-1*%pi)
 -1.020646306179612b0*cos(2.5b-1*%pi)

(%i5) bfloat(%);
(%o5) 4.741476366409942b-1*%i-9.692642119442159b-1

------------------------------------
Is this necessity to sometimes use two applications of bfloat to reduce an
erf expression to a x.xxby form a result of the definition of the erf 
function
or a result of how the bigfloat package processes any expression?

Ted

p.s. I was originally trying to get
       bfloat (integrate (sin(x)/sqrt(x), x, 0, 1)) reduced.