Matplotlib colours are wrong - python-3.x

Question

NOTE: This is now resolved, although I made no changes to my code, but the images are now mysteriously coming out with perfect colour.
I am making plots using Python3 with a Spyder interface. I have looked at the plots both in the Spyder terminal, and in the saved PNGs of the images. Both show that the colours are coming out very badly.
A simple line in the script
plt.plot(x1, beta.transpose()[x1], color = 'r', linewidth = 0.2)
The red colour is showing as a light pink. When I change 'r' for 'k' to get a black line, it comes as a very light grey. Does anyone know of issues affecting Matplotlib, or Matplotlib used through Spyder, that might account for this?
Example:
Code:
Data-generating script logit_data.py
#Logistic regression data and link function.
import numpy as np
#Define link function here
def g(z):
g=1/(1+np.exp(-z))
return g
#For producing y data values given true paramters theta and number of covariates
def logit_data(n,p, theta):
#Define parameters
#1)Number of covariates
p_i = p+1 #with intercept
p_i=np.int(p_i)
#2) m as correct data type
n=np.int(n)
#4)Specify parameter valueas to be estimated
theta=np.reshape(theta, (p_i,1))
#5)Define distribution from which covariate values are drawn i.i.d., and initiate data values
X=np.zeros((n,p_i))
X[:,0]=1 #intercept
mean=0
sigma=1.5
X[:,1:]=np.random.normal(mean,sigma,(n,p))
#6)Produce y values treating y as a Bernoulli variable with p=g(X*theta)
r=np.random.uniform(0,1,n)
r=np.reshape(r, (len(r),1))
htrue=g(X.dot(theta))
y=htrue-r
y[y>=0]=1
y[y<0]=0
return X, y
Plotting script:
#Script for producing y data from p covariates from a specified distribution, specified beta paraemters,
#and n data samples for logit link function.
import numpy as np
import matplotlib.pyplot as plt
from logit_data import logit_data
import pylab
import statsmodels.stats as sms
import statsmodels.api as sma
import csv
def figure2():
#def MLE_logistic_function():
#1)Sample and observation numbers
samples = 30
observations = 40000
#2)Number of independent covariates
p=299
#3)True beta to be estimated (parameter values)
nonzerosN=30
beta1=np.append(np.full((1, nonzerosN),10),np.full((1,nonzerosN),-10), axis=1)
print(np.shape(beta1))
beta=np.append(beta1,np.zeros((1,p+1-2*nonzerosN)), axis=1)
print(np.shape(beta))
#4)#Initiate arrays to store estimates of beta (and errors) computed at specified sample numbers N
#Betas=np.zeros((len(npowers),p+1))
#Errors=np.zeros((len(npowers),p+1))
#5)Obtain random covariate values from specified distribution, and corresponding y values using true beta
#plus gaussian noise term.
X,y = logit_data(observations,p,beta)
logit = sma.Logit(y,X)
result = logit.fit()
print(result.summary())
MLEcoefficients = result.params
x1 = np.arange(0, nonzerosN,1)
x2 = np.arange(nonzerosN, 2*nonzerosN,1)
x3 = np.arange(2*nonzerosN, p+1,1)
plt.scatter(index, MLEcoefficients, 0.2)
plt.plot(x1, beta.transpose()[x1], color = 'black', linewidth = 0.2, alpha=1)
plt.plot(x2, beta.transpose()[x2], color = 'black', linewidth = 0.2)
plt.plot(x3, beta.transpose()[x3], color = 'black', linewidth = 0.2)
plt.xlabel('Index')
plt.ylabel('Coefficient values (true and fitted)')
plt.savefig('MMLTfig2_p%s_o%d.png' %(p,observations))
plt.show()
return
figure2()