Print the Capitals without the Capitals

Mathematica, 168 153 149 bytes - 20% = 119.2 bytes

u="\.55nited\.53tates";\.41dministrative\.44ivision\.44ata[{#,u}&/@\.43ountry\.44ata[u,"\.52egions"],{"\.43apital\.4eame","\.53tate\.41bbreviation"}]

Obligatory, but I did not know that any character can be replaced by \.xx or \:xxxx with the appropriate hex code.

Edit: Cut of 4 more characters by replacing Thread with a pure function.

Output:

{{Montgomery,AL},{Juneau,AK},{Phoenix,AZ},{Little Rock,AR},{Sacramento,CA},{Denver,CO},{Hartford,CT},{Dover,DE},{Washington,DC},{Tallahassee,FL},{Atlanta,GA},{Honolulu,HI},{Boise,ID},{Springfield,IL},{Indianapolis,IN},{Des Moines,IA},{Topeka,KS},{Frankfort,KY},{Baton Rouge,LA},{Augusta,ME},{Annapolis,MD},{Boston,MA},{Lansing,MI},{Saint Paul,MN},{Jackson,MS},{Jefferson City,MO},{Helena,MT},{Lincoln,NE},{Carson City,NV},{Concord,NH},{Trenton,NJ},{Santa Fe,NM},{Albany,NY},{Raleigh,NC},{Bismarck,ND},{Columbus,OH},{Oklahoma City,OK},{Salem,OR},{Harrisburg,PA},{Providence,RI},{Columbia,SC},{Pierre,SD},{Nashville,TN},{Austin,TX},{Salt Lake City,UT},{Montpelier,VT},{Richmond,VA},{Olympia,WA},{Charleston,WV},{Madison,WI},{Cheyenne,WY}}

CJam, 312 bytes

".ýç9.5i-jæ¤þ¸«Ã«cj­|ù;ÎüÄ`­Ñ¯Äÿçsøi4ÔÚ0;¾o'ÈàÚãÕ»®¼v{Ðù·*ñfiö\^é]ù¬ðö¸qÚpÿ©a$ÿÆhk¥½éØ×ïÕ{ñ9ÁÛ%Ðø¦ð·âßxâj   Ö묭¯,Ð+?Û¡!ù%Âí©Úfx`¤|}¼>qñµÉÎ4Óæj-wöÄÆ 4,üÖáÌxsÍ·üãýÛêmÁj±æ0?³¢¶§%Û57Ëmc.~`b=´á¥ÉpË,ôb¶ÌsÁì¾*§òÿ_Ö©;<tíèz6ljç¸b§èäø>`ÍÚפÒòÔ§~hÝ®Ú8¼}8Ì7rÿé×ÔÎîæ¡©)Ô@"'[fm256,f=)b27b'`f+'`/{_2>'q/32af.^' *o2<eup}/

The code is 390 bytes long and qualifies for the 20% bonus.

Note that the code is riddled with unprintable characters. Try it online in the CJam interpreter.

Idea

We have to encode the output somehow without using uppercase letters or the numbers 1 and 65 to 90 anywhere in the code.

We start by rewriting the desired output as

akjuneau`paharrisburg`txaustin`maboston`wvcharleston`azphoenix`kyfrankfort`msjackson`mdannapolis`vtmontpelier`ndbismarck`hihonolulu`meaugusta`nvcarsonqcity`sccolumbia`ohcolumbus`wycheyenne`casacramento`arlittleqrock`nmsantaqfe`mnsaintqpaul`idboise`tnnashville`codenver`nhconcord`almontgomery`inindianapolis`riprovidence`utsaltqlakeqcity`ilspringfield`ncraleigh`labatonqrouge`sdpierre`dedover`orsalem`waolympia`kstopeka`varichmond`cthartford`nyalbany`milansing`njtrenton`mthelena`iadesqmoines`gaatlanta`wimadison`nelincoln`fltallahassee`okoklahomaqcity`mojefferson

By subtracting the character ` from all characters of that string, we obtain an array containing integers from 0 to 26. We can convert this array from base 27 to base 229, yielding an array of integers from 0 to 228.

If we add 91 to each base-229 digit and take the results modulo 256, we map the range [0, …, 164] to [91, … 255], and the range [165, …, 228] to [0, …, 63]. This leaves the characters with code points from 64 (@) to 90 (Z) unused.

The string to encode is not in the same order as the sample output in the question. I tried several permutation until I found one that contains no null bytes, linefeeds, carriage returns or non-breaking spaces (problematic with the online interpreter), and no double quotes (require escaping).

Code

"…@"   e# Push a string of 342 ISO-8859-1 characters.
'[fm   e# Subtract the char '[' (code point 91) from each char of the string.
256,f= e# Compute the remainder of the differences divided by 256.
)      e# Pop the last integer from the array ('@' -> 27 -> 229).
b27b   e# Convert the remaining array from base 229 to base 27.
'`f+   e# Add the character '`' to the resulting digits.
       e# This pushes the string from the "Idea" section.
'`/    e# Split the result at backticks.
{      e# For each resulting chunk C:
  _2>  e#   Copy C and remove its first to characters.
  'q/  e#   Split at occurrences of 'q'.
  32a  e#   Push [32].
  f.^  e#   Mapped, vectorized XOR; XOR the first character of each chunk 
       e#   with 32. This changes its case.
  ' *  e#   Join the resulting chunks, separating by spaces.
  o    e#   Print.
  2<   e#   Reduce the original C to its first two characters.
  eu   e#   Convert to uppercase.
  p    e#   Print, enclosed in double quotes, and followed by a linefeed.
}/     e#

R, 96 bytes 98 bytes -20% -> 78.4

Thanks to @plasticinsect for the bonus!

library(maps);data(us.cities);cat(gsub("()( \\w+)$",",\\2",us.cities$n[us.cities$ca==2]),sep="\n")

Previous code at 96 bytes:

library(maps);data(us.cities);cat(gsub("( \\w+)$",",\\1",us.cities$n[us.cities$ca==2]),sep="\n")

From package maps, it loads a dataset of US cities. Column capital contains a 2 if it is a state capital. The names of the cities are given is column name in the form "City StateAbbreviation" (i. e. Albany NY), so one need to add an explicit delimiter in between before output. To do so I eventually use the regex \1 meaning I can't have the bonus I suppose. To avoid using \1 in the regex I added an empty group so that I can use \2.

Usage:

> library(maps);data(us.cities);cat(gsub("()( \\w+)$",",\\2",us.cities$n[us.cities$ca>1]),sep="\n")
Albany, NY
Annapolis, MD
Atlanta, GA
Augusta, ME
Austin, TX
Baton Rouge, LA
Bismarck, ND
Boise, ID
Boston, MA
Carson City, NV
Charleston, WV
Cheyenne, WY
Columbia, SC
Columbus, OH
Concord, NH
Denver, CO
Des Moines, IA
Dover, DE
Frankfort, KY
Harrisburg, PA
Hartford, CT
Helena, MT
Honolulu, HI
Indianapolis, IN
Jackson, MS
Jefferson City, MO
Juneau, AK
Lansing, MI
Lincoln, NE
Little Rock, AR
Madison, WI
Montgomery, AL
Montpelier, VT
Nashville, TN
Oklahoma City, OK
Olympia, WA
Phoenix, AZ
Pierre, SD
Providence, RI
Raleigh, NC
Richmond, VA
Sacramento, CA
Saint Paul, MN
Salem, OR
Salt Lake City, UT
Santa Fe, NM
Springfield, IL
Tallahassee, FL
Topeka, KS
Trenton, NJ