The problem is that you fix the width of the two figures to be the same in gridspec and then add the colorbar to the second plot which takes up some of the width (5% plus the padding). And as imshow has an 1:1 aspect ration this change of width also changes the height of the figure.
The simplest way to avoid that is to account for the additional width of the colorbar in the width ratio:
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
from mpl_toolkits.axes_grid1 import make_axes_locatable
# Random test data:
ROI = np.random.rand(250, 150)
z = np.random.rand(250, 150)
fig = plt.figure(num=2,figsize=(4,2))
gs = gridspec.GridSpec(1, 2, width_ratios=[1,1.08])
ax0 = plt.subplot(gs[0])
ax1 = plt.subplot(gs[1])
ax0.imshow(ROI,'gray')
map1 = ax1.imshow(z,cmap=plt.cm.YlOrRd)
divider1 = make_axes_locatable(ax1)
cax1 = divider1.append_axes("right", size="5%", pad="3%")
cbar1 = plt.colorbar(map1, cax=cax1)
plt.show()
This will result in the following plot:

You can also avoid that by simply adding a third axes for the colorbar using gridspec:
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
# Random test data:
ROI = np.random.rand(250, 150)
z = np.random.rand(250, 150)
fig = plt.figure(num=2,figsize=(5,2))
gs = gridspec.GridSpec(1, 3, width_ratios=[1,1,0.05])
ax0 = plt.subplot(gs[0])
ax1 = plt.subplot(gs[1])
cax1 = plt.subplot(gs[2])
ax0.imshow(ROI,'gray')
map1 = ax1.imshow(z,cmap=plt.cm.YlOrRd)
cbar1 = plt.colorbar(map1, cax=cax1)
plt.show()
This gives:

This is probably more flexible but also requires more tweaking of padding and sizes on your part.