Why TextRecognize reads data organized in grid by column and not by row?

Question

Why TextRecognize reads the text by column and not by line in the following scenarios?

The lines are perfect, the font is perfect, image quality is perfect so I do not understand why it does not read it normally by line.

Also notice that it reads 66 as 55, in fact every 6 it read as 5. Also in the second example it reads UJ as U] or XB as xB or QC as Qc.

The conditions can not be more perfect so why it fails in such ideal situation?

Input:

im=ImagePad[Rasterize[Grid[Partition[ToString[#[[1]]]~StringJoin~ToString[#[[2]]]&/@Partition[RandomInteger[9,105],2],7,7,{1,1},{}]],ImageSize->300],10,White]
TextRecognize[im,"Line"]

Output:

{"58", "95", "08", "55", "25", "02", "49", "84", "13", "04", "18", \
"74", "78", "19", "25", "95", "95", "35", "59", "00", "18", "94", \
"49", "10", "20", "32", "82", "88", "32", "49", "25", "98", "70", \
"87", "59", "28", "27", "84", "10", "41", "32", "27", "55", "25", \
"23", "57", "00", "55", "83", "42", "80", "97"}

Input:

im=ImagePad[Rasterize[Grid[Partition[ToString[#[[1]]]~StringJoin~ToString[#[[2]]]&/@Partition[CharacterRange["A","Z"][[1+RandomInteger[25,105]]],2],7,7,{1,1},{}]],ImageSize->300],10,White]
TextRecognize[im,"Line"]

Output:

{"BF", "LS", "IZ", "IX", "ZP", "NO", "ME", "QL", "LF", "YL", "SO", \
"LE", "FV", "CS", "EX", "CT", "FY", "BX", "KL", "EU", "KS", "WR", \
"LT", "PA", "CX", "IK", "PI", "GV", "MN", "OT", "FK", "xB", "DY", \
"Qc", "DN", "UL", "HB", "wz", "FY", "MH", "U]", "PK", "XL", "YT", \
"PQ", "DM", "IP", "CS", "SM", "Z]", "WV", "KF"}