\begin{table}%t3
\caption{\label{tab:curvedspectra} Special fit results.}
%%\centerline

{\small
\begin{tabular}{llccccccc}
\hline\hline\\[-3mm]
Data Set & Used Fit & $f_0$ & $\alpha^{(')}$ & $\beta$ & $E_{\rm cut}$ [TeV] & $\chi^2_{\rm red, fit}$ & Likelihood  & $E_{\rm peak}$ [TeV] \\
\hline 
\multirow{4}*{April 27, 2006}		& PL      & $9.54\pm0.52$ & $1.92\pm0.07$ & & & \phantom{0}5.3/4 && \\
           & log-$P$ & $9.35\pm0.55$ & $1.54\pm0.19$ & $0.59\pm0.29$ & & 0.48/3 & $96\%$ & \phantom{0} $1.2\pm 0.7$ \phantom{0}\\						& log-$P$ apex & $11.5\pm0.9\phantom{0}$ & $0.26\pm0.17$ & &  & 0.48/3 & $96\%$ & $1.2\pm0.2$ \\
& PL+C & $11.3\pm1.2 \phantom{0}$ & $1.44\pm0.24$ && $2.6\pm1.3$ & 0.34/3 & $96\%$ & \phantom{0} $1.4\pm 1.0$ \phantom{0} \\\hline
\multirow{4}*{All April Data}		& PL      & $4.53\pm0.07$ & $2.07\pm0.04$ & & & \phantom{0} 16/5&& \\
           & log-$P$ & $4.75\pm0.12$ & $1.89\pm0.06$ & $0.39\pm0.11$ & & \phantom{0}1.2/4 & $99\%$ & $0.69\pm 0.14$ \\
	& log-$P$ apex &  $4.84\pm0.16$ & $0.41\pm0.11$ & & & \phantom{0}1.2/4 & $99\%$ & $0.69\pm0.06$ \\
	& PL+C &  $5.36\pm0.31$ & $1.77\pm0.09$ & & $3.6\pm1.1$ & \phantom{0}1.8/4 & $99\%$&  \phantom{0} $0.80\pm 0.42$  \phantom{0} \\\hline
\multirow{4}*{High-State Nights}	& PL      &  $8.19\pm0.28$ & $1.93\pm0.05$ & & & \phantom{0}6.0/4 && \\
           & log-$P$ &  $8.46\pm0.32$ & $1.79\pm0.09$ & $0.29\pm0.15$ & & \phantom{0}2.0/3 & $94\%$ & \phantom{0} $1.1\pm 0.6$ \phantom{0} \\
	& log-$P$ apex &  $9.21\pm0.47$ & $0.52\pm0.17$ & &  & \phantom{0}2.0/3 & $94\%$ &  $1.1\pm0.3$ \\
& PL+C &  $9.02\pm0.64$ & $1.75\pm0.12$ & & $6.1\pm4.0$ & \phantom{0}3.1/3 & $91\%$& \phantom{0} $1.5\pm 1.2$ \phantom{0}\\\hline
\multirow{4}*{Low-State Nights}	& PL      &  $3.39\pm0.10$ & $2.17\pm0.05$ & & & \phantom{0}6.6/4 && \\
           & log-$P$ &  $3.55\pm0.13$ & $2.02\pm0.08$ & $0.38\pm0.17$ & & \phantom{0}1.1/3 & $97\%$ & $0.48\pm 0.12$ \\
	& log-$P$ apex &  $3.55\pm0.13$ & $0.41\pm0.16$ & & & \phantom{0}1.1/3 & $97\%$&  $0.48\pm0.12$  \\
	& PL+C &  $4.15\pm0.40$ & $1.85\pm0.15$ & & $2.9\pm1.3$  & 0.75/3 & $97\%$ & $0.45\pm 0.47$ \\
\hline
\end{tabular}}
\tablefoot{Results of a power-law fit (Eq.~(\ref{eq:2})), a log-parabolic fit (Eq.~(\ref{eq:4})  $\times$ $(E/E_0)^2$), a log-parabolic fit in apex form (Eq.~(\ref{eq:apex})) and
a power-law fit with an exponential cutoff ((Eq.~(\ref{eqn:PL+C}) $\times$
$(E/E_0)^2$) in $E^2 {\rm d}F/{\rm d}E$ after EBL de-absorption for
special data sets. $f_0$ is given in units of $10^{-11}~{\rm TeV}^{-1}~{\rm cm}^{-2}~ {\rm s}^{-1}$; $\alpha$, $\alpha'$, $\beta$, $E_{\rm cut}$ and $E_{\rm peak}$ are the fit parameters as stated in the text, and Likelihood denotes the probability of a likelihood ratio test. The on/off normalization factor is $1/3$, $E_0=0.5$~TeV.}
\end{table}